# PlungeAI > Developer documentation for the PlungeAI One API, the MCP server and the agent skills. - [PlungeAI API Overview](https://docs-preview.plungeai.com/getting-started/overview.md): Connect PlungeAI to your coding agent, onboard it, and build on the One API and the MCP server. - [Choose an API](https://docs-preview.plungeai.com/getting-started/choose-an-api.md): Find the right PlungeAI API for what you are building. - [PlungeAI API Pricing](https://docs-preview.plungeai.com/getting-started/pricing.md): How PlungeAI meters and bills model calls and runs. (planned) - [API Rate Limits](https://docs-preview.plungeai.com/getting-started/rate-limits.md): Default limits: execution planes 30 to 300 per minute by tier, models plane 600 per minute per key. - [Your first calls](https://docs-preview.plungeai.com/getting-started/quickstart.md): One key, one agent call, one model call, one MCP connection. - [Authentication & keys](https://docs-preview.plungeai.com/getting-started/authentication.md): The three key prefixes, which routes accept which, and how to keep keys safe. - [How PlungeAI works](https://docs-preview.plungeai.com/getting-started/concepts.md): The seven planes, the MCP endpoint and how a request flows. - [Models Plane Quickstart](https://docs-preview.plungeai.com/models/quickstart.md): OpenAI-compatible chat completions with one `sk-ocean-` key and a base-URL swap. (planned) - [Models Best Practices](https://docs-preview.plungeai.com/models/best-practices.md): Request fields, tools and structured output, and what each provider supports. (planned) - [Evaluating Models](https://docs-preview.plungeai.com/models/evaluating-models.md): Planned: compare models on your own gold set, with cost and latency. (planned) - [Routing: models, sort, failover](https://docs-preview.plungeai.com/models/routing.md): Ordered fallback with `models[]`, `sort` by price, latency or throughput, and presets. (planned) - [Migrate to PlungeAI](https://docs-preview.plungeai.com/models/migrate.md): Move from OpenRouter, the OpenAI SDK, LiteLLM or the Vercel AI SDK. (planned) - [Advanced Models Settings](https://docs-preview.plungeai.com/models/advanced-settings.md): Guardrails, the opt-in response cache and BYOK-or-billed routing. (planned) - [Upgrade from guide 2.0 to 3.0](https://docs-preview.plungeai.com/models/migration-guide.md): What changed between the 2.0 and 3.0 developer guides, for every plane. (planned) - [Embeddings Quickstart](https://docs-preview.plungeai.com/models/embeddings.md): OpenAI-compatible embeddings (`openai/*` models only). (planned) - [Models, providers and slugs](https://docs-preview.plungeai.com/models/catalog.md): Read the live catalog with `GET /v1/models` and pick a model id. (planned) - [Models Billing & Metering](https://docs-preview.plungeai.com/models/billing.md): Account status, price = cost × org markup, reserve-then-settle, spend caps. (planned) - [Migrate from the legacy SDKs](https://docs-preview.plungeai.com/models/sdk-migration.md): Move from `@ocean-platform/external-sdk` and the Python, Go and Rust SDKs. (planned) - [Models Streaming](https://docs-preview.plungeai.com/models/streaming.md): SSE chat completions, translated for non-OpenAI providers. (planned) - [Workflows Quickstart](https://docs-preview.plungeai.com/workflows/quickstart.md): Run multi-agent CNL workflows inline or saved, sync or streamed. (planned) - [Workflow Best Practices](https://docs-preview.plungeai.com/workflows/best-practices.md): Write CNL that validates, runs, and fails loudly. (planned) - [The CNL workflow spec](https://docs-preview.plungeai.com/workflows/core-concepts/cnl.md): Workflow structure, task types and validation rules. (planned) - [Choosing agents and models](https://docs-preview.plungeai.com/workflows/core-concepts/agents-and-models.md): Agent tiers, model pinning, and how much time parallel fan-out takes. (planned) - [Workflow Runs Lifecycle](https://docs-preview.plungeai.com/workflows/core-concepts/run-lifecycle.md): Create a run, read its status and result, cancel it. (planned) - [Results and traces](https://docs-preview.plungeai.com/workflows/core-concepts/results-and-traces.md): What a run returns and how to read its persisted trace. (planned) - [Follow-ups and continuation](https://docs-preview.plungeai.com/workflows/core-concepts/follow-ups.md): Chain a run with follow-ups and answer paused runs. (planned) - [Interactive Chat](https://docs-preview.plungeai.com/workflows/examples/chat.md): A multi-turn chat session over MCP. (planned) - [Deep Research](https://docs-preview.plungeai.com/workflows/examples/deep-research.md): A goal-driven research workflow that answers with sources. (planned) - [Batch Enrichment](https://docs-preview.plungeai.com/workflows/examples/enrichment.md): Enrich a list of items with a `type: batch` workflow, then run it to completion. (planned) - [Workflow Webhooks](https://docs-preview.plungeai.com/workflows/webhooks.md): Planned: get notified when a run finishes. (planned) - [Parallel and batch tasks](https://docs-preview.plungeai.com/workflows/parallel-tasks.md): Fan out with `type: parallel` and run a per-item pipeline with `type: batch`. (planned) - [Generate a workflow from a goal](https://docs-preview.plungeai.com/workflows/generate.md): `plungeai_build_workflow` turns a goal into CNL YAML. (planned) - [Workflow streaming events](https://docs-preview.plungeai.com/workflows/streaming-events.md): SSE engine frames from `execute-stream`. (planned) - [MCP servers inside workflows](https://docs-preview.plungeai.com/workflows/mcp-tool-calling.md): Give a harness task MCP servers to call. (planned) - [Saved workflows and versions](https://docs-preview.plungeai.com/workflows/saved-workflows.md): Save, version and run workflows by id. (planned) - [Agents Quickstart](https://docs-preview.plungeai.com/agents/quickstart.md): Execute a registry agent from a prompt: sync, async or streamed. (planned) - [Conversations and personas](https://docs-preview.plungeai.com/agents/features/conversations.md): Carry context with `messages[]` (partial: `prompt` stays required and threading depends on the agent), `persona`, and MCP `session_id` for reliable threading. (planned) - [Output formats](https://docs-preview.plungeai.com/agents/features/output-formats.md): JSON, YAML, Markdown or text via `format` and `Accept`. (planned) - [Agent streaming](https://docs-preview.plungeai.com/agents/features/streaming.md): OpenAI-shaped chunks with keepalives and `X-Execution-Id`. (planned) - [Citations](https://docs-preview.plungeai.com/agents/features/citations.md): Planned: source annotations on agent answers. (planned) - [Web search tool agents](https://docs-preview.plungeai.com/agents/features/web-search.md): Search the web through registry tool agents. (planned) - [Outbound MCP runs](https://docs-preview.plungeai.com/agents/features/mcp-tools.md): Connect catalog MCP servers to a run and call their tools. (planned) - [Async execution and results](https://docs-preview.plungeai.com/agents/features/async-results.md): `sync:false` returns 202; redeem the result later. (planned) - [Direct requests](https://docs-preview.plungeai.com/agents/examples/direct-requests.md): Ask a question, ask for JSON, send YAML and get Markdown. (planned) - [Use an agent as a subagent tool](https://docs-preview.plungeai.com/agents/examples/subagent.md): Wrap an agent call as a tool in your own orchestrator. (planned) - [OpenAI compatibility (agents plane)](https://docs-preview.plungeai.com/agents/openai-compatibility.md): Which OpenAI request and stream fields the agents plane accepts. (planned) - [Discovery Quickstart](https://docs-preview.plungeai.com/discovery/quickstart.md): Find agents, tools, models, skills and connectors before you call. (planned) - [Recommend](https://docs-preview.plungeai.com/discovery/recommend.md): Ask the registry which capability fits a task. (planned) - [Search modes and fields](https://docs-preview.plungeai.com/discovery/core-concepts/search-modes.md): `mode` hybrid, keyword or vector; `fields` list, summary or full. (planned) - [The card object](https://docs-preview.plungeai.com/discovery/core-concepts/cards.md): What a registry card holds and how to read one. (planned) - [Card status and kinds](https://docs-preview.plungeai.com/discovery/core-concepts/card-status.md): Only `status:active` cards run; kinds you can filter on. (planned) - [Inspect a tool contract](https://docs-preview.plungeai.com/discovery/features/tool-contract.md): Read operations, required params and examples before executing. (planned) - [Quality signals](https://docs-preview.plungeai.com/discovery/features/quality.md): Add `include=quality` to search results. (planned) - [Catalog resources over MCP](https://docs-preview.plungeai.com/discovery/features/resources.md): Dropped: discovery search is synchronous and has no event stream. Replacement: read catalog lists as MCP resources. (planned) - [Registry kinds](https://docs-preview.plungeai.com/discovery/features/kinds.md): Dropped: there is no long-running discovery run to notify on. Replacement: the kinds you can search. (planned) - [Pagination](https://docs-preview.plungeai.com/discovery/features/pagination.md): Discovery search and `GET /v1/tools` page with `limit` and `offset` and return `count`; only `GET /v1/agents` returns `has_more` and `next_offset`. (planned) - [Workflow templates](https://docs-preview.plungeai.com/discovery/features/templates.md): Dropped: a search cannot be cancelled. Replacement: start from a template. (planned) - [Discovery first](https://docs-preview.plungeai.com/discovery/features/discovery-first.md): Dropped: there is no stateful FindAll-style run to refresh. Replacement: the discovery-first rule. (planned) - [Scheduling Quickstart](https://docs-preview.plungeai.com/scheduling/quickstart.md): Cron jobs for agents, queries, workflows and heartbeats. (planned) - [Run now and run history](https://docs-preview.plungeai.com/scheduling/runs.md): Trigger a job now and read its run history. (planned) - [Webhook triggers](https://docs-preview.plungeai.com/scheduling/webhook-triggers.md): Planned: start a job from an inbound webhook on a public host. (planned) - [Heartbeat check](https://docs-preview.plungeai.com/scheduling/examples/heartbeat.md): A heartbeat job that checks a condition and acts on it. (planned) - [Trigger a workflow or mission](https://docs-preview.plungeai.com/scheduling/examples/trigger-workflow.md): Run a workflow or mission on a schedule and notify a channel. (planned) - [Scheduling: MCP tool → REST API](https://docs-preview.plungeai.com/scheduling/migration-guide.md): Planned: move jobs from `plungeai_schedule` to a REST schedules API. (planned) - [Tools Quickstart](https://docs-preview.plungeai.com/tools/quickstart.md): List tool agents, read a contract, execute one operation. (planned) - [Outcomes and remediation](https://docs-preview.plungeai.com/tools/outcomes.md): The seven outcomes, the remediation actions, and their HTTP statuses. (planned) - [Tool catalog by category](https://docs-preview.plungeai.com/tools/catalog.md): The tool agents available today, by category. (planned) - [Long runs (async)](https://docs-preview.plungeai.com/tools/async.md): `mode: "async"` for tool calls that take longer than one request. (planned) - [Missions Quickstart](https://docs-preview.plungeai.com/missions/quickstart.md): Run a bounded agent loop with a goal, allowed tools and an iteration cap. (planned) - [Mission bounds and the tool fence](https://docs-preview.plungeai.com/missions/bounds.md): What a mission may do and where it stops. (planned) - [Skills, plugins and capability injection](https://docs-preview.plungeai.com/missions/capability-injection.md): Inject skills, experts, persona, backgrounds, plugins and MCP. (planned) - [Pre-built agent cards](https://docs-preview.plungeai.com/missions/prebuilt-agents.md): Run a catalog agent card with `mission_ref` or `pack`. (planned) - [Async missions: the Ten-Minute Relay](https://docs-preview.plungeai.com/missions/async-relay.md): Long missions run in resumable legs; poll for the result. (planned) - [Connectors and connected accounts](https://docs-preview.plungeai.com/connectors/quickstart.md): OAuth and API-key connectors, and how a connected account reaches a call. (planned) - [The connector proxy](https://docs-preview.plungeai.com/connectors/proxy.md): Call a provider's API raw through `gateway.plungeai.com` with an `sk-conn-` key (`$PLUNGE_CONNECTOR_KEY`). (planned) - [Traces Quickstart](https://docs-preview.plungeai.com/traces/quickstart.md): Read the spans and gateway log of any run by trace id. (planned) - [Execution history and cost](https://docs-preview.plungeai.com/traces/history-and-cost.md): List past runs and what they cost. (planned) - [The activity log](https://docs-preview.plungeai.com/traces/activity-log.md): Your key's call log on the MCP server. (planned) - [Developer Tools Overview](https://docs-preview.plungeai.com/developer-tools/quickstart.md): MCP vs One API vs CLI vs Studio: pick your door. (planned) - [Account API](https://docs-preview.plungeai.com/developer-tools/account-api.md): Planned: manage keys, wallet and usage with a key. Today these live in the Dashboard. (planned) - [Install PlungeAI Agent Skills](https://docs-preview.plungeai.com/developer-tools/agent-skills.md): 31 skills that teach coding agents to use PlungeAI. (planned) - [Anthropic Tool Calling](https://docs-preview.plungeai.com/developer-tools/anthropic-tool-calling.md): Turn PlungeAI tool contracts into Claude tool definitions. (planned) - [Claude Code Plugin](https://docs-preview.plungeai.com/developer-tools/claude-code-plugin.md): The `plungeai` plugin: 31 skills plus its own MCP server config. Install it instead of `claude mcp add`, not in addition. (planned) - [Ocean CLI](https://docs-preview.plungeai.com/developer-tools/cli.md): The `ocean` command: auth, runs, workflows, schedules, registry. (planned) - [Cursor Plugin](https://docs-preview.plungeai.com/developer-tools/cursor-plugin.md): Planned: a Cursor marketplace plugin. (planned) - [LangChain](https://docs-preview.plungeai.com/developer-tools/langchain.md): Use the models plane from LangChain's `ChatOpenAI`. (planned) - [PlungeAI MCP Server Quickstart](https://docs-preview.plungeai.com/developer-tools/mcp/quickstart.md): Connect any MCP client to `https://mcp.plungeai.com/v1`. - [Programmatic use](https://docs-preview.plungeai.com/developer-tools/mcp/programmatic-use.md): Call the MCP server from your own code with raw JSON-RPC, and run outbound MCP servers with `/v1/mcp/runs`. (planned) - [PlungeAI MCP](https://docs-preview.plungeai.com/developer-tools/mcp/plungeai-mcp.md): The one server: install per client, configure, troubleshoot. (planned) - [Workflows and missions over MCP](https://docs-preview.plungeai.com/developer-tools/mcp/workflows-mcp.md): Run workflows and missions from your coding agent, sync or async. (planned) - [Protocol and transport](https://docs-preview.plungeai.com/developer-tools/mcp/protocol.md): Streamable HTTP, JSON-RPC methods, sessions, SSE progress. (planned) - [Prompts and resources](https://docs-preview.plungeai.com/developer-tools/mcp/prompts-resources.md): The `/plungeai` prompt and every `plungeai://` resource. (planned) - [Move MCP clients from /mcp to /v1](https://docs-preview.plungeai.com/developer-tools/mcp/migrate-to-v1.md): `/mcp` is a legacy alias since 2026-09-20; configure `/v1`. (planned) - [Docs MCP server](https://docs-preview.plungeai.com/developer-tools/mcp/docs-mcp.md): Planned: search and read these docs from your agent. (planned) - [OAuth Provider](https://docs-preview.plungeai.com/developer-tools/oauth-provider.md): Planned: sign in with PlungeAI for MCP clients that require OAuth. Today: bearer `ozk_` keys only. (planned) - [Ollama Tool Calling](https://docs-preview.plungeai.com/developer-tools/ollama-tool-calling.md): Turn PlungeAI tool contracts into tool definitions for a local Ollama model through its OpenAI-compatible endpoint. (planned) - [OpenAI SDK](https://docs-preview.plungeai.com/developer-tools/openai-sdk.md): Point the OpenAI Python or TypeScript SDK at PlungeAI. (planned) - [OpenAI Tool Calling](https://docs-preview.plungeai.com/developer-tools/openai-tool-calling.md): Turn tool contracts into OpenAI function definitions and execute the calls. (planned) - [Generate a client from OpenAPI](https://docs-preview.plungeai.com/developer-tools/openapi-codegen.md): Typed clients from `/v1/openapi.json`. (planned) - [OpenCode Plugin](https://docs-preview.plungeai.com/developer-tools/opencode-plugin.md): Connect OpenCode today with an MCP config; a plugin is planned. (planned) - [@plungeai/one-api](https://docs-preview.plungeai.com/sdks/typescript.md): Planned: the typed TypeScript client for the execution planes. Today: generate a client from the spec. (planned) - [PlungeAI in Claude Code](https://docs-preview.plungeai.com/integrations/claude-code.md): One `claude mcp add` command, or a project `.mcp.json`. (planned) - [PlungeAI in Claude.ai and Claude Desktop](https://docs-preview.plungeai.com/integrations/claude-ai.md): Skill upload on claude.ai; mcp-remote bridge on Desktop. (planned) - [PlungeAI in Codex](https://docs-preview.plungeai.com/integrations/codex.md): `codex mcp add` with a bearer-token env var. (planned) - [PlungeAI in Cursor](https://docs-preview.plungeai.com/integrations/cursor.md): One-click deeplink or `~/.cursor/mcp.json`. (planned) - [PlungeAI in VS Code](https://docs-preview.plungeai.com/integrations/vscode.md): One-click install or `code --add-mcp`. (planned) - [PlungeAI in Gemini CLI](https://docs-preview.plungeai.com/integrations/gemini-cli.md): `~/.gemini/settings.json` with `httpUrl`. (planned) - [PlungeAI in OpenCode](https://docs-preview.plungeai.com/integrations/opencode.md): `opencode.json` remote MCP entry. (planned) - [PlungeAI in Windsurf](https://docs-preview.plungeai.com/integrations/windsurf.md): `~/.codeium/windsurf/mcp_config.json` snippet. (planned) - [PlungeAI in Replit](https://docs-preview.plungeai.com/integrations/replit.md): One-click install into Replit Agent. (planned) - [PlungeAI in Lovable](https://docs-preview.plungeai.com/integrations/lovable.md): Chat connector or app connector. (planned) - [PlungeAI in Bolt.new](https://docs-preview.plungeai.com/integrations/bolt.md): Custom MCP server in Connectors. (planned) - [PlungeAI in v0](https://docs-preview.plungeai.com/integrations/v0.md): Add MCP in the prompt form; the generated app cannot call MCP. (planned) - [Any other MCP client](https://docs-preview.plungeai.com/integrations/other-mcp-clients.md): URL, header and transport for clients not listed here. (planned) - [Agentic payments (x402)](https://docs-preview.plungeai.com/integrations/agentic-payments.md): Pay x402 services in USDC through the payment agents, with approval. (planned) - [AWS Marketplace](https://docs-preview.plungeai.com/integrations/aws-marketplace.md): Planned: subscribe through AWS Marketplace. (planned) - [Browser Use](https://docs-preview.plungeai.com/integrations/browser-use.md): Planned: private-web data through a browser agent. (planned) - [Google Cloud Marketplace](https://docs-preview.plungeai.com/integrations/google-cloud-marketplace.md): Planned: subscribe through Google Cloud Marketplace. (planned) - [Google Gemini Enterprise](https://docs-preview.plungeai.com/integrations/google-gemini-enterprise.md): Planned: PlungeAI as a grounding source in Gemini Enterprise. (planned) - [Google Sheets](https://docs-preview.plungeai.com/integrations/google-sheets.md): Planned: a sheet function that runs a PlungeAI agent. (planned) - [LiteLLM](https://docs-preview.plungeai.com/integrations/litellm.md): Use the models plane as an OpenAI-wire LiteLLM provider. (planned) - [n8n](https://docs-preview.plungeai.com/integrations/n8n.md): Planned: n8n nodes for agents, tools and workflows. (planned) - [OpenRouter compatibility](https://docs-preview.plungeai.com/integrations/openrouter.md): Endpoint and field parity with OpenRouter, and what you gain. (planned) - [Render](https://docs-preview.plungeai.com/integrations/render.md): Planned: a one-click deploy template for a sample app. (planned) - [Superhuman](https://docs-preview.plungeai.com/integrations/superhuman.md): Planned: PlungeAI skills in Superhuman. (planned) - [Vercel](https://docs-preview.plungeai.com/integrations/vercel.md): Vercel AI SDK against the models plane; v0 over MCP. (planned) - [Zapier](https://docs-preview.plungeai.com/integrations/zapier.md): Planned: a Zapier app for agents and workflows. (planned) - [Data Integrations](https://docs-preview.plungeai.com/data-integrations/overview.md): Planned: run PlungeAI agents over rows in your data tools. (planned) - [Apache Spark](https://docs-preview.plungeai.com/data-integrations/spark.md): Planned: SQL-native UDFs that call agents. (planned) - [DuckDB](https://docs-preview.plungeai.com/data-integrations/duckdb.md): Planned: batch enrichment from DuckDB. (planned) - [Google BigQuery](https://docs-preview.plungeai.com/data-integrations/bigquery.md): Planned: BigQuery remote functions. (planned) - [Polars](https://docs-preview.plungeai.com/data-integrations/polars.md): Planned: a Polars expression plugin. (planned) - [Snowflake](https://docs-preview.plungeai.com/data-integrations/snowflake.md): Planned: a Snowflake UDTF. (planned) - [Supabase](https://docs-preview.plungeai.com/data-integrations/supabase.md): Planned: Supabase Edge Function helpers. (planned) - [PlungeAI API Glossary](https://docs-preview.plungeai.com/resources/glossary.md): Every term used across the API, MCP and skills docs. (planned) - [Connector catalog](https://docs-preview.plungeai.com/resources/connector-catalog.md): OAuth and API-key connectors, BYOK, and how connected accounts reach calls. (planned) - [Memory](https://docs-preview.plungeai.com/resources/memory.md): Long-term memory, and distilling sessions into skills. (planned) - [Roles and permissions](https://docs-preview.plungeai.com/resources/roles-and-permissions.md): Company and team roles, scopes and levels in the Dashboard. (planned) - [Trust fence and approvals](https://docs-preview.plungeai.com/resources/trust-fence.md): Key fences, the trust fence, the agent fence, mission bounds and model guardrails. (planned) - [API Error Codes and Warnings](https://docs-preview.plungeai.com/resources/warnings-and-errors.md): Every error envelope and code, the outcomes, and what to retry. (planned) - [Webhook setup](https://docs-preview.plungeai.com/resources/webhook-setup.md): Planned: signing, retries and verification for outbound webhooks. (planned) - [Migrate from legacy execute aliases](https://docs-preview.plungeai.com/workflows/migration-guide.md): Move `/v1/execute` and `/v1/cnl/*` calls to `/v1/workflows/*`. (planned) - [Docs for AI agents](https://docs-preview.plungeai.com/resources/docs-for-agents.md): Dropped: PlungeAI runs no web crawler, so webmaster guidance does not apply. Replacement: how agents read these docs (`/agents.md`, llms.txt, `.md` twins, Docs MCP). - [Status](https://docs-preview.plungeai.com/resources/status.md): Planned: a status page. For now, the health endpoints. (planned) - [FAQs](https://docs-preview.plungeai.com/resources/faqs.md): Answers to common platform, API, billing and security questions. (planned) - [Create chat completion](https://docs-preview.plungeai.com/api-reference/models/create-chat-completion.md): Send messages to a model and get a completion, or stream it as SSE chunks. Ordered fallback with `models[]`, `sort`, presets and an opt-in response cache. - [Create embeddings](https://docs-preview.plungeai.com/api-reference/models/create-embeddings.md): Turn a string or a list of strings into embedding vectors (OpenAI-compatible). - [List models](https://docs-preview.plungeai.com/api-reference/models/list-models.md): The priced model catalog routing candidates are drawn from. `pricing.in` and `pricing.out` are the provider cost basis in USD per 1M tokens, before your org markup. - [Search the registry](https://docs-preview.plungeai.com/api-reference/discovery/search-registry.md): Hybrid semantic and keyword search over capability cards. Pages with `limit` and `offset`; a page shorter than `limit` is the last one. - [Recommend a card](https://docs-preview.plungeai.com/api-reference/discovery/recommend-card.md): Get the one best capability card for a task, with a score. - [Get a card](https://docs-preview.plungeai.com/api-reference/discovery/get-card.md): Read one capability card as markdown (the LLM view): operations, examples and "Not for" redirects. - [List tools](https://docs-preview.plungeai.com/api-reference/tools/list-tools.md): List the active tool agents, 50 per page by default (at most 100). - [Get a tool contract](https://docs-preview.plungeai.com/api-reference/tools/get-tool-contract.md): The invocation contract of one tool agent: JSON Schema parameters, operations, approval gates and examples. - [Execute a tool](https://docs-preview.plungeai.com/api-reference/tools/execute-tool.md): Run one operation of a tool agent with typed params. Gated and money verbs refuse unattended calls. - [List agents](https://docs-preview.plungeai.com/api-reference/agents/list-agents.md): List the active agents, 50 per page by default (at most 100). - [List agent categories](https://docs-preview.plungeai.com/api-reference/agents/list-agent-categories.md): Agent categories with the number of active agents in each. - [Execute an agent](https://docs-preview.plungeai.com/api-reference/agents/execute-agent.md): Run one agent with a prompt. Sync by default; `sync: false` returns a pointer to redeem later; `stream: true` streams OpenAI-shaped chunks. - [Get an agent result](https://docs-preview.plungeai.com/api-reference/agents/get-agent-result.md): Redeem the result of an async agent run with the `workflow_id` and `task_id` it returned. - [Execute an inline workflow](https://docs-preview.plungeai.com/api-reference/workflows/execute-inline-workflow.md): Run a CNL workflow sent in the body, as JSON `{workflow, input}` or as a raw `text/yaml` document. Returns a pointer; read the content from the result route. - [Stream an inline workflow](https://docs-preview.plungeai.com/api-reference/workflows/stream-inline-workflow.md): Run an inline workflow and stream every engine event as SSE until `workflow_completed` or `workflow_error`. - [Execute a saved workflow](https://docs-preview.plungeai.com/api-reference/workflows/execute-saved-workflow.md): Run one of your saved workflows by id with an optional `input` or `inputs`. - [Stream a saved workflow](https://docs-preview.plungeai.com/api-reference/workflows/stream-saved-workflow.md): Run a saved workflow and stream its engine events as SSE. - [Get a workflow result](https://docs-preview.plungeai.com/api-reference/workflows/get-workflow-result.md): Read the stored output of one task of a run, usually the `final_task_id` the execute call returned. - [Cancel an execution](https://docs-preview.plungeai.com/api-reference/workflows/cancel-execution.md): Ask a running execution of yours to stop. Cancellation is cooperative. - [List MCP tools](https://docs-preview.plungeai.com/api-reference/mcp/list-mcp-tools.md): The MCP server tool list as a plain GET; the answer is the JSON-RPC `tools/list` envelope with the tools under `result.tools`. - [MCP endpoint (JSON-RPC)](https://docs-preview.plungeai.com/api-reference/mcp/mcp-endpoint.md): The Streamable HTTP MCP endpoint: `initialize`, then `tools/list` and `tools/call`. The samples use the MCP host `https://mcp.plungeai.com/v1`, the same server. - [Open an MCP run](https://docs-preview.plungeai.com/api-reference/mcp/create-mcp-run.md): Connect catalog MCP servers for a run and get their namespaced tool table. Always close the run when done. - [List MCP run tools](https://docs-preview.plungeai.com/api-reference/mcp/list-mcp-run-tools.md): List the namespaced tools (`mcp____`) of an open MCP run. - [Call an MCP run tool](https://docs-preview.plungeai.com/api-reference/mcp/call-mcp-run-tool.md): Call one namespaced tool of an open MCP run with its arguments. - [Close an MCP run](https://docs-preview.plungeai.com/api-reference/mcp/close-mcp-run.md): Close an MCP run and drop its state. Idempotent. - [Get a trace](https://docs-preview.plungeai.com/api-reference/traces/get-trace.md): The persisted trace of one run: execution spans and the gateway request log for the trace id. - [Health](https://docs-preview.plungeai.com/api-reference/meta/health.md): Liveness probe of the One API router. No key needed. - [OpenAPI spec](https://docs-preview.plungeai.com/api-reference/meta/get-openapi-spec.md): The OpenAPI 3.1 document of the One API. The docs copy with tags is at [/openapi.json](/openapi.json). - [llms.txt](https://docs-preview.plungeai.com/api-reference/meta/llms-txt.md): The One API guide index for AI coding agents, as plain text (llmstxt.org). - [llms-full.txt](https://docs-preview.plungeai.com/api-reference/meta/llms-full-txt.md): The whole One API Developer Guide 3.0 and the external API reference in one markdown file. - [Guide (HTML)](https://docs-preview.plungeai.com/api-reference/meta/docs.md): The One API Developer Guide 3.0 as one HTML page on the API host. - [Resolve a task to one tool, its operation and an example call, with a…](https://docs-preview.plungeai.com/api-reference/other/post-v1-discovery-resolve.md): Resolve a task to one tool, its operation and an example call, with a calibrated confidence - [MCP server overview](https://docs-preview.plungeai.com/mcp-reference/overview.md): Endpoint, versions, auth and sessions. - [plungeai_execute_workflow](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_execute_workflow.md): Execute a CNL workflow (saved workflow_id or ad-hoc workflow_yaml). - [plungeai_execute_agent](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_execute_agent.md): Execute a single PROMPT-DRIVEN agent (e.g. - [plungeai_get_tool_contract](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_get_tool_contract.md): The exact invocation contract for one registry agent: JSON Schema for its parameters, operations (with approval gates), worked YAML examples, output shape, and LIVE credential status for the acting user (platform-managed vs "connect Google first"). - [plungeai_execute_tool](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_execute_tool.md): Execute one structured tool-agent directly with typed params: {agent_id, operation?, params}. - [plungeai_run_mission](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_run_mission.md): Run a bounded autonomous agent mission (loop agent): give a goal, optional tool fence and iteration cap. - [plungeai_learn](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_learn.md): Manage your private skill library. - [plungeai_list_agents](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_list_agents.md): Discover platform building-block agents from the live registry (the capability catalog). - [plungeai_list_workflows](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_list_workflows.md): List your saved workflows — what users usually mean by "my agents". No args shows your folders (rubrics) + recent; filter with folder and/or search. - [plungeai_get_result](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_get_result.md): Retrieve the result of one of your executions (by `execution_id`) — for conversational runs this includes the FULL conversation thread (all follow-up turns), same as Studio shows. - [plungeai_get_workflow_status](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_get_workflow_status.md): Check the status of one of your executions (self-heals stuck runs). Returns structured content too, including `continuation` when the run is paused awaiting user approval or an answer — relay that to the user and use plungeai_continue. - [plungeai_workflow](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_workflow.md): Manage your workflows: action create|get|update|delete|save_version|list_versions|get_version|restore_version. Create/update/delete sync live to Studio, iMini, and peer apps. - [plungeai_executions](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_executions.md): Browse your execution history: action list|get|output|conversation|delete|cancel. - [plungeai_followup](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_followup.md): Ask a follow-up about a completed execution (continues its agent session or researches with prior context). If the execution is still running, your message is queued and injected at its next turn (steer); if finished, it continues the conversation. - [plungeai_continue](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_continue.md): Continue a paused conversation: answer a question (message) or approve a pending action (approve:true). Only pass approve:true after the user explicitly confirmed — never approve on your own. To deny or change course, pass their words as message. - [plungeai_chat](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_chat.md): Persistent chat with the PlungeAI assistant (platform tools + web search). action send|new|list_sessions|history. Conversations appear in Studio and the CLI. - [plungeai_build_workflow](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_build_workflow.md): Generate a new workflow from a goal, or refine an existing one (workflow_id + instruction), via the platform builder. Saves it so it appears everywhere (live sync). - [plungeai_memory](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_memory.md): Long-term memory scoped to you: action recall|remember|search_runs|get_run. - [plungeai_templates](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_templates.md): Workflow templates: action list|get|use. "use" creates a workflow from the template (optional name + folder). - [plungeai_schedule](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_schedule.md): Scheduled (cron) jobs: action stats|list|get|create|update|pause|resume|delete|run_now|runs. create needs name, job_type (agent|query|workflow|heartbeat), target, schedule; create also takes mission_ref (schedule a pre-built agent card). - [plungeai_whoami](https://docs-preview.plungeai.com/mcp-reference/tools/plungeai_whoami.md): Show the authenticated identity: user id, tier, key label, rate-limit window. - [PlungeAI Agents](https://docs-preview.plungeai.com/mcp-reference/resources/agents-list.md): Live agent registry (markdown) - [Agent Categories](https://docs-preview.plungeai.com/mcp-reference/resources/agents-categories.md): Live registry categories with agent counts (markdown) - [Personas](https://docs-preview.plungeai.com/mcp-reference/resources/personas-list.md): Live persona registry (markdown) - [Your Workflows](https://docs-preview.plungeai.com/mcp-reference/resources/workflows-list.md): Your saved workflows (markdown) - [Workflow Authoring Guide](https://docs-preview.plungeai.com/mcp-reference/resources/docs-workflow-authoring.md): Author, validate, and save CNL workflows (full skill bundle) - [Agents by category](https://docs-preview.plungeai.com/mcp-reference/resources/agents-by-category.md): Resource template plungeai://agents/{category}, read with resources/read. - [Workflow YAML by id](https://docs-preview.plungeai.com/mcp-reference/resources/workflow-yaml.md): Resource template plungeai://workflows/{id}, read with resources/read. - [Execution result by id](https://docs-preview.plungeai.com/mcp-reference/resources/execution-result.md): Resource template plungeai://executions/{id}, read with resources/read. - [plungeai](https://docs-preview.plungeai.com/mcp-reference/prompts/plungeai.md): The /plungeai prompt: PlungeAI MCP. - [initialize](https://docs-preview.plungeai.com/mcp-reference/protocol/initialize.md): Negotiate the protocol version and read the server capabilities. - [tools/list](https://docs-preview.plungeai.com/mcp-reference/protocol/tools-list.md): List the tools your key can call; there is no pagination. - [tools/call](https://docs-preview.plungeai.com/mcp-reference/protocol/tools-call.md): Call one tool; the answer streams as SSE. - [resources/list](https://docs-preview.plungeai.com/mcp-reference/protocol/resources-list.md): List the five static resources. - [resources/templates/list](https://docs-preview.plungeai.com/mcp-reference/protocol/resources-templates-list.md): List the three resource templates. - [resources/read](https://docs-preview.plungeai.com/mcp-reference/protocol/resources-read.md): Read one resource by URI; an unknown URI answers -32002. - [prompts/list](https://docs-preview.plungeai.com/mcp-reference/protocol/prompts-list.md): List the one prompt, plungeai. - [prompts/get](https://docs-preview.plungeai.com/mcp-reference/protocol/prompts-get.md): Get a prompt by name; an unknown name answers -32602. - [ping](https://docs-preview.plungeai.com/mcp-reference/protocol/ping.md): Check the connection; the result is an empty object. - [logging/setLevel](https://docs-preview.plungeai.com/mcp-reference/protocol/logging-set-level.md): Accepted and answered with an empty object; the level is not kept. - [Get balance](https://docs-preview.plungeai.com/account-api/balance/get-balance.md): Dashboard view: the wallet balance, top-ups, spend and auto-recharge setting. (planned) - [Add to balance](https://docs-preview.plungeai.com/account-api/balance/add-to-balance.md): Dashboard view: top up the wallet through Stripe Checkout. (planned) - [Create key](https://docs-preview.plungeai.com/account-api/keys/create-key.md): Create an ozk_ key in the Dashboard. (planned) - [Delete key](https://docs-preview.plungeai.com/account-api/keys/delete-key.md): Dashboard view: revoke a key in the Dashboard. (planned) - [Get usage](https://docs-preview.plungeai.com/account-api/usage/get-usage.md): Dashboard view: runs, tokens and cost by user, workflow, model or channel. (planned) - [Who am I](https://docs-preview.plungeai.com/account-api/identity/whoami.md): Your identity, tier, key label and rate-limit window over MCP with plungeai_whoami. - [Agent Skills](https://docs-preview.plungeai.com/skills.md): 31 skills, version 3.0.0 - [choose-your-plungeai-door](https://docs-preview.plungeai.com/skills/choose-your-plungeai-door.md): Picks the right PlungeAI entry point before you write anything — the MCP server (https://mcp.plungeai.com/v1) for live agent operation, the One API (https://api.plungeai.com) for code you write or generate, the Ocean CLI for a terminal or CI workflow, or Ocean Studio for a human building visually… - [plungeai-api-setup](https://docs-preview.plungeai.com/skills/plungeai-api-setup.md): Setup for the PlungeAI One API (https://api.plungeai.com) — a self-service ozk_ key, the base URL, the Bearer/X-API-Key auth header, your first curl call, the live OpenAPI 3.1 contract at /v1/openapi.json, typed-client codegen (openapi-typescript, openapi-python-client, the future… - [SDK & codegen — typed clients, snippets, and the AGENTS.md block](https://docs-preview.plungeai.com/skills/plungeai-api-setup/references/sdk-and-codegen.md): The supported integration stance, in order of preference: - [plungeai-cli-setup](https://docs-preview.plungeai.com/skills/plungeai-cli-setup.md): Operate the Ocean CLI (`ocean`; package @plungeai/ocean-cli — not yet on npm, run from the Ocean-Platform repo) to drive PlungeAI from a terminal: install/setup, auth (Studio session cookie vs self-service ozk_ key and which commands need which), running saved workflows and ad-hoc CNL YAML, async… - [Ocean CLI — agent flows, REPL, and scripting](https://docs-preview.plungeai.com/skills/plungeai-cli-setup/references/agent-flows.md): All execution happens on Cloudflare; the CLI streams or polls remotely. - [Ocean CLI — install, auth, and setup](https://docs-preview.plungeai.com/skills/plungeai-cli-setup/references/auth-and-setup.md): The npm package @plungeai/ocean-cli (v2.3.0, bin name ocean, Node >= 20) is not yet published — publishing is an owner action (pushing the ocean-cli-v2.3.0 tag triggers the publish workflow). Current truth from cli/README.md: - [Ocean CLI — full command reference](https://docs-preview.plungeai.com/skills/plungeai-cli-setup/references/commands.md): Every command below exists in cli/src/index.ts (Commander registration) and is implemented under cli/src/commands/. Global flags: -h/--help on any command, -V/--version on the root. Running ocean with no arguments opens the AI chat REPL. - [plungeai-mcp-setup](https://docs-preview.plungeai.com/skills/plungeai-mcp-setup.md): Connect and operate the PlungeAI MCP server (https://mcp.plungeai.com/v1) — endpoint and Bearer ozk_/X-API-Key auth, a generic MCP client config (native remote/HTTP, or the mcp-remote stdio bridge), plungeai_whoami, the structured-outcome envelope and trust fences, the full plungeai_* tool index… - [Identity & errors — whoami, the auth model, outcome envelope, trust fences](https://docs-preview.plungeai.com/skills/plungeai-mcp-setup/references/identity-and-errors.md): --- - [plungeai-agents](https://docs-preview.plungeai.com/skills/plungeai-agents.md): Run a PlungeAI registry agent: recognize the two kinds (prompt-driven vs structured tool-agents), execute a prompt-driven one sync or async (`plungeai_execute_agent` MCP / `POST /v1/agents/{id}/execute` REST, incl. - [Executing agents over MCP — execute_agent, get_result](https://docs-preview.plungeai.com/skills/plungeai-agents/references/mcp-execute-and-result.md): Two execution styles, chosen by the agent's card: - [Agents — the registry, discovery, and execution](https://docs-preview.plungeai.com/skills/plungeai-agents/references/registry.md): An agent is a deployed capability with a uniform interface: it accepts a task, does one job well (search the web, call an LLM, convert a document, post to Slack, enrich a CRM contact), and stores its result where the platform can hand it to the next step. - [Agents plane — /v1/agents](https://docs-preview.plungeai.com/skills/plungeai-agents/references/rest-agents-plane.md): Execute any active platform agent with a plain prompt. The gateway wraps your call in a one-task workflow, so every execution yields a workflowid + taskid pair — that pair is the handle for async redemption and for traces. - [plungeai-campaigns](https://docs-preview.plungeai.com/skills/plungeai-campaigns.md): Run a list to completion on PlungeAI: the campaign ledger (claim/complete/fail/retry over data-table-agent), the campaign-config block (list source, cycle, cadence, retries, delivery), and how a campaign keeps going via cron or self-scheduling runAgain. - [Campaigns — running a list to completion (the campaign ledger)](https://docs-preview.plungeai.com/skills/plungeai-campaigns/references/campaigns.md): A campaign claims a list of items (a data-table query, an agent's lister output, an uploaded CSV, or an inline list), runs each item through a CNL pipeline with per-item retries, and repeats on a cadence until the list ("cycle") is drained — then either re-cycles (refill/reset) or hands off to… - [plungeai-discovery](https://docs-preview.plungeai.com/skills/plungeai-discovery.md): Find the right thing on PlungeAI's live registry before building anything: agents, structured tools, models, skills, personas, connectors, and saved workflow templates, via hybrid semantic + keyword search (`plungeai_list_agents` MCP / `GET /v1/discovery/search` REST), full capability cards, a… - [Discovery over MCP — list_agents, get_tool_contract, templates (listing)](https://docs-preview.plungeai.com/skills/plungeai-discovery/references/list-and-contract.md): The registry is LIVE: agents come from the registry service at request time, no static manifest. - [Discovery — /v1/discovery](https://docs-preview.plungeai.com/skills/plungeai-discovery/references/search-and-recommend.md): Discovery is the live catalog of everything the platform can do — agents, workflows, MCP servers, connectors, models — with hybrid semantic + keyword search ranked server-side. - [plungeai-memory](https://docs-preview.plungeai.com/skills/plungeai-memory.md): Read and write PlungeAI's per-user long-term memory (plungeai_memory: recall/remember/search_runs/get_run) and distill a session into a reusable skill with plungeai_learn — distinct from SharedMemory (a single run's output). - [Memory — run data (SharedMemory) vs long-term memory. Two systems, never confuse them](https://docs-preview.plungeai.com/skills/plungeai-memory/references/memory.md): PlungeAI has TWO memory systems with different jobs: - [plungeai-missions](https://docs-preview.plungeai.com/skills/plungeai-missions.md): Run bounded autonomous PlungeAI agent missions (type: harness) — a goal, a tool fence, an iteration cap, and self-checked success criteria, via plungeai_run_mission or a harness workflow task. - [Missions — bounded autonomous agent runs (type: harness)](https://docs-preview.plungeai.com/skills/plungeai-missions/references/missions.md): A mission bounds what an agent may do for ONE goal: a standing purpose, a fenced tool set, an iteration cap, optional identity/knowledge injection, and self-checked success criteria. - [plungeai-models](https://docs-preview.plungeai.com/skills/plungeai-models.md): Model routing on PlungeAI: how agents/workflows/missions resolve a model through the platform's provider factory (`model`/`provider` fields on a task or mission), and the separate OpenAI-compatible money-plane API (`https://api.plungeai.com/v1/chat/completions`, `/v1/embeddings`, `/v1/models`)… - [Models plane — /v1/chat/completions, /v1/embeddings, /v1/models](https://docs-preview.plungeai.com/skills/plungeai-models/references/money-plane.md): The OpenAI-compatible surface ("money plane"). Auth is a different key from the rest of the API: Authorization: Bearer sk-ocean-YOURKEY (from Ocean Dashboard → One API → Keys, "Model gateway keys" panel). Errors use the OpenAI error shape {"error":{"message","type","code","requestid"}}. - [Models — provider factory inside runs, money plane for your code](https://docs-preview.plungeai.com/skills/plungeai-models/references/provider-factory.md): PlungeAI touches models in two distinct places. Keep them apart: - [plungeai-platform](https://docs-preview.plungeai.com/skills/plungeai-platform.md): THE PlungeAI (Ocean) platform capability map: a growing live registry of agents and structured tools, CNL multi-agent workflows, bounded harness missions, injectable skills/plugins/experts/personas, model routing with an OpenAI-compatible inference API, cron scheduling, per-user long-term memory… - [plungeai-results-traces](https://docs-preview.plungeai.com/skills/plungeai-results-traces.md): Debug and observe PlungeAI runs: live SSE events, plungeai_get_workflow_status / plungeai_executions history, persisted GET /v1/traces/{id} spans and gateway request cost, the HITL conversation loop (plungeai_continue/plungeai_followup/plungeai_chat), and outbound MCP door runs (POST /v1/mcp/runs). - [Conversation — chat, followup, continue (and the ⏸ HITL loop)](https://docs-preview.plungeai.com/skills/plungeai-results-traces/references/conversation.md): Three distinct conversation surfaces — pick by what the user is doing: - [MCP plane — /v1/mcp](https://docs-preview.plungeai.com/skills/plungeai-results-traces/references/mcp-runs.md): Two directions in one plane: - [Observability — watching, debugging, and pricing platform runs](https://docs-preview.plungeai.com/skills/plungeai-results-traces/references/observability.md): Every execution on PlungeAI is observable at four altitudes: live events while it runs, an execution record after it finishes, a persisted span-level trace, and the gateway's own request log with cost. Debugging is reading these in order — not re-running the workload to "see what happens". - [Traces — GET /v1/traces/{id}](https://docs-preview.plungeai.com/skills/plungeai-results-traces/references/traces.md): The observability plane: one id follows a request across every internal hop. (This is the traces half of the One API's discovery+traces route group — for /v1/discovery/ see plungeai-discovery.) - [plungeai-scheduling](https://docs-preview.plungeai.com/skills/plungeai-scheduling.md): Cron-schedule PlungeAI workflows, agent queries, pre-built agent cards, and condition-watching heartbeats via plungeai_schedule, with full run history and retries. - [Scheduling — cron jobs for workflows, agents, and missions](https://docs-preview.plungeai.com/skills/plungeai-scheduling/references/scheduling.md): The scheduler runs platform work unattended on cron expressions: a saved workflow every morning, an agent query hourly, a pre-built agent card nightly. - [plungeai-skills-plugins](https://docs-preview.plungeai.com/skills/plungeai-skills-plugins.md): Declare and understand PlungeAI's capability-injection fields on a type: harness task or plungeai_run_mission — skills, experts, persona, backgrounds, plugins, and MCP servers — including eager-vs-deferred budgets and degrade-to-warning failure semantics. - [Experts, personas, backgrounds — shaping WHO the agent is in a run](https://docs-preview.plungeai.com/skills/plungeai-skills-plugins/references/experts-personas.md): Three capability kinds shape an agent's identity and framing (as opposed to skills, which shape its methods). All three inject into the front of the system prompt, in a fixed order, before anything else the run adds: - [Plugins — bundles of skills, MCP servers, and scripts injected as one unit](https://docs-preview.plungeai.com/skills/plungeai-skills-plugins/references/plugins.md): A plugin is a Claude-Code-format bundle stored on the platform: a directory tree containing any number of SKILL.md skills, an optional .mcp.json (or manifest mcpServers) documenting the MCP servers its skills were written for, and executable Python scripts. - [Skills — injectable instruction packs for agent runs](https://docs-preview.plungeai.com/skills/plungeai-skills-plugins/references/skills.md): A skill is a markdown instruction pack (the Agent Skills SKILL.md format: frontmatter name + description, then the body) stored on the platform and injected into agent runs. - [plungeai-tools-connectors](https://docs-preview.plungeai.com/skills/plungeai-tools-connectors.md): Call PlungeAI structured tool-agents with typed parameters against a published invocation contract: fetch the contract (`plungeai_get_tool_contract` / `GET /v1/tools/{id}`), execute with typed `params` (`plungeai_execute_tool` / `POST /v1/tools/{id}/execute`), read live… - [Tools — structured tool-agents and their contracts](https://docs-preview.plungeai.com/skills/plungeai-tools-connectors/references/contract-and-execution.md): A tool on PlungeAI is a structured agent: instead of free-text prompts it takes typed parameters against a published contract — named operations, a JSON Schema for inputs, worked examples, and per-operation approval gates. - [Executing tools over MCP — execute_tool](https://docs-preview.plungeai.com/skills/plungeai-tools-connectors/references/mcp-execute-tool.md): Structured tool-agents (cards with a Parameters table — posting, document conversion, weather, data-table CRUD, calendar operations, payment actions) take typed fields against a published contract, not prose. - [Tools plane — /v1/tools](https://docs-preview.plungeai.com/skills/plungeai-tools-connectors/references/rest-tools-plane.md): Every active agent is also callable as a tool with a machine-checkable invocation contract. The flow is always: discover → inspect the contract → execute. This plane runs the platform's trust fence — gated and money operations refuse unattended execution by design. - [plungeai-workflows](https://docs-preview.plungeai.com/skills/plungeai-workflows.md): Build, validate, test, and save PlungeAI (Ocean Studio) workflows in CNL YAML. - [Workflows plane — /v1/workflows](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/api.md): Run multi-agent CNL workflows: inline (send the workflow document in the request), saved (reference a workflow id saved in Ocean Studio), each with an optional SSE streaming variant, plus a results-redemption route. - [Getting a workflow into PlungeAI — three channels](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/channels.md): Connect once (self-service ozk key from Dashboard → One API → Keys, https://dashboard.plungeai.com): - [CNL Workflow Reference (engine v6) — PlungeAI](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/cnl-spec.md): CNL (Cognitive Natural Language) is PlungeAI's YAML workflow language: you declare tasks and how they compose (parallel, sequential, batch, debate, harness, …), and the engine dispatches them to platform agents with automatic data flow between steps. - [Workflows — execute_workflow, get_workflow_status, list_workflows, workflow, build_workflow, executions](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/mcp.md): Workflows are CNL YAML run by the platform engine. - [Workflows — CNL orchestration: what it is and how to run it](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/overview.md): A workflow is a YAML document (CNL — Cognitive Natural Language) that the engine executes as a DAG of agent calls: parallel fan-out, sequential pipelines, unlimited nesting, conditions, batches, debates, and bounded autonomous missions. - [CNL recipes — pick the shape, then adapt](https://docs-preview.plungeai.com/skills/plungeai-workflows/references/recipes.md): Full runnable versions live in examples/. Two composition rules apply to all of them: - [plungeai-in-bolt](https://docs-preview.plungeai.com/skills/plungeai-in-bolt.md): Connect Bolt.new to PlungeAI (Ocean Studio) over MCP — the Connectors → Custom MCP server form, the all-tools/all-projects toggle behavior, keeping the shipped app's key server-side, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-claude-ai](https://docs-preview.plungeai.com/skills/plungeai-in-claude-ai.md): Connect claude.ai web or Claude Desktop to PlungeAI (Ocean Studio) — Desktop live tools via the mcp-remote bridge in claude_desktop_config.json (OAuth isn't available yet, hence the bridge), claude.ai web via skill.zip upload, and a plungeai_whoami + plungeai_list_agents verify for Desktop. - [plungeai-in-claude-code](https://docs-preview.plungeai.com/skills/plungeai-in-claude-code.md): Connect the Claude Code CLI to PlungeAI (Ocean Studio) over MCP — 'claude mcp add' one-liner, project-scoped .mcp.json with an env-var key, the skills/plugin install path, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-codex](https://docs-preview.plungeai.com/skills/plungeai-in-codex.md): Connect OpenAI Codex CLI to PlungeAI (Ocean Studio) over MCP — native streamable-HTTP in ~/.codex/config.toml (env-var or static bearer token), the codex mcp add terminal command, the mcp-remote bridge fallback, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-cursor](https://docs-preview.plungeai.com/skills/plungeai-in-cursor.md): Connect Cursor to PlungeAI (Ocean Studio) over MCP — the one-click install-page deeplink or a ~/.cursor/mcp.json entry, Agent/Plan-mode tool behavior, the ~40-tool cap, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-gemini-cli](https://docs-preview.plungeai.com/skills/plungeai-in-gemini-cli.md): Connect Gemini CLI to PlungeAI (Ocean Studio) over MCP — a ~/.gemini/settings.json entry using httpUrl (not url, which is SSE-only and 405s), the mcp-remote bridge fallback, /mcp reload after edits, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-lovable](https://docs-preview.plungeai.com/skills/plungeai-in-lovable.md): Connect Lovable to PlungeAI (Ocean Studio) — a personal chat connector (any plan) so Lovable's chat can operate PlungeAI via MCP while building, and a workspace-admin app connector so shipped apps call the One API in production with attached auth, plus a plungeai_whoami + plungeai_list_agents… - [plungeai-in-opencode](https://docs-preview.plungeai.com/skills/plungeai-in-opencode.md): Connect OpenCode to PlungeAI (Ocean Studio) over MCP — an opencode.json mcp.plungeai entry with type: remote, {env:VAR} variable substitution for a committed project config, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-replit](https://docs-preview.plungeai.com/skills/plungeai-in-replit.md): Connect Replit Agent to PlungeAI (Ocean Studio) over MCP — the one-click install-page link, the Integrations → MCP Servers form, keeping the Agent's MCP key separate from a deployed app's Replit Secret, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-v0](https://docs-preview.plungeai.com/skills/plungeai-in-v0.md): Connect v0 (Vercel) to PlungeAI (Ocean Studio) over MCP — the + menu → MCPs form, why the generated Next.js app must call PlungeAI through a server Route Handler (v0 apps can't call MCP directly), and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-vscode](https://docs-preview.plungeai.com/skills/plungeai-in-vscode.md): Connect VS Code (GitHub Copilot agent mode) to PlungeAI (Ocean Studio) over MCP — the code --add-mcp CLI command, a committed-safe .vscode/mcp.json with an input-prompted key, the 128-tools-per-request cap, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-in-windsurf](https://docs-preview.plungeai.com/skills/plungeai-in-windsurf.md): Connect Windsurf (Cascade) to PlungeAI (Ocean Studio) over MCP — a ~/.codeium/windsurf/mcp_config.json entry using serverUrl, the mcp-remote bridge fallback for stdio-only builds, the 100-tool cap, and a plungeai_whoami + plungeai_list_agents verify. - [plungeai-agentic-agent](https://docs-preview.plungeai.com/skills/plungeai-agentic-agent.md): Design and emit ONE bounded agentic agent for PlungeAI (Ocean Studio) as a single `type: harness` mission — purpose, allowed_tools fence, permissions gates (deny/ask), skills/plugins/MCP/persona declarations, effort or turn cap, success criteria — then hand it to Studio as CNL YAML. - [The fence — tools, agents, and fan-out guards](https://docs-preview.plungeai.com/skills/plungeai-agentic-agent/references/fence.md): allowedtools is a hard whitelist — only listed tools ever reach the LLM; an out-of-fence call is refused by the runtime, not merely discouraged. Omit it and the loop runtime applies its own fail-closed default fence. - [Harness task — the full mission contract](https://docs-preview.plungeai.com/skills/plungeai-agentic-agent/references/harness-task.md): In a saved agentic agent: memoryowner is forbidden (tenancy comes from the executing user); local / localagents are bot-only and require agent: harness-agent. - [plungeai-bot-agent](https://docs-preview.plungeai.com/skills/plungeai-bot-agent.md): Design and emit ONE PlungeAI (Ocean Studio) bot agent — a scheduled, unattended `type: harness` mission that reports to the user's channels (in-app, email, Slack, Telegram, WhatsApp, Discord) — as CNL YAML plus a bot-config block (cron schedule, delivery targets, optional run-on-my-computer). - [bot-config → scheduler job](https://docs-preview.plungeai.com/skills/plungeai-bot-agent/references/bot-config.md): Studio turns the block into one scheduled job: jobtype: workflow, target: , schedule: , parameters (JSON) { prompt, deliver, local }. The scheduler runs the saved workflow at each tick and delivers taskcomplete's result to every target. - [plungeai-campaign-agent](https://docs-preview.plungeai.com/skills/plungeai-campaign-agent.md): Design and emit ONE PlungeAI (Ocean Studio) campaign agent — a long-running, list-driven, resumable agent that works an owned ledger of items in short scheduled runs until the current cycle is exhausted — as CNL YAML (a claim task plus a `type: batch` with `ledger: campaign`) and a campaign-config… - [campaign-config reference](https://docs-preview.plungeai.com/skills/plungeai-campaign-agent/references/campaign-config.md): A campaign-config fenced block sits after the CNL YAML. Studio parses it to open the ledger and create the scheduler job. Fields: - [Changelog](https://docs-preview.plungeai.com/resources/changelog.md): Planned: the dated list of PlungeAI changes. (planned) ## OpenAPI Specs - [openapi](https://docs-preview.plungeai.com/openapi.json) - [api-live-openapi](https://api.plungeai.com/v1/openapi.json) - [mcp-tools](https://docs-preview.plungeai.com/mcp-tools.json) ## Optional - [Get API Key](https://dashboard.plungeai.com/one-api?tab=keys) - [OpenAPI Spec](https://docs-preview.plungeai.com/openapi.json) - [TypeScript SDK](https://docs-preview.plungeai.com/sdks/typescript) - [Ocean CLI](https://docs-preview.plungeai.com/developer-tools/cli)