# Pharlo

> Pharlo is a content posting API and MCP server for publishing, scheduling, and tracking social media content. Live platforms today: YouTube and Facebook.

Pharlo lets AI agents and applications publish posts, schedule content, manage connected channels, and read stats across social platforms. Two integration surfaces share one backend:

- MCP server (for AI agents): OAuth 2.0 Bearer auth, 21 tools.
- REST API (for developers): API key auth (`ds_live_...`), OpenAPI 3.0.

Live platforms are YouTube and Facebook only. Other platforms may appear in listings but are not yet publishable.

## MCP server
- [MCP endpoint](https://pharlo.io/_mcp): Streamable HTTP MCP endpoint. Authenticate with OAuth 2.0 (Bearer token).
- [MCP tools catalog](https://pharlo.io/mcp-tools.json): Machine-readable catalog of all 21 MCP tools — names, input schemas, behavior annotations, credit costs, when-to-use guidance, and worked examples.
- [MCP tools reference](https://pharlo.io/docs/mcp/tools): Human-readable reference for all 21 MCP tools — when to use each, inputs, outputs, examples, and recovery flows.
- [Agent workflows](https://pharlo.io/docs/agent-workflows): Task-oriented playbooks that map real user intents to concrete tool sequences with preconditions and recovery paths.
- [Agent rules](https://pharlo.io/docs/agent-rules): Safety and correctness rules for AI agents acting on a user's behalf — publishing, scheduling, media, failure handling, and secrets.
- [MCP overview](https://pharlo.io/mcp/): What the MCP server does and how to connect it.
- [MCP discovery manifest](https://pharlo.io/.well-known/mcp.json): Service-discovery manifest — server URL, transport, auth metadata, and supported platforms.
- [MCP client compatibility](https://pharlo.io/docs/mcp/compatibility): Which MCP clients work with the Pharlo server and what has been tested — all use OAuth 2.0 Bearer over Streamable HTTP.

## REST API
- [OpenAPI spec (JSON)](https://pharlo.io/openapi.json): Canonical OpenAPI 3.0 specification for the REST API.
- [OpenAPI spec (YAML)](https://pharlo.io/openapi.yaml): The same specification in YAML.
- [API reference](https://pharlo.io/api/doc): Interactive Swagger UI for the REST API.

## Docs
- [Full operational reference](https://pharlo.io/llms-full.txt): Single-file compact reference consolidating tools, platforms, workflows, errors, billing, and safety rules for AI ingestion.
- [Documentation](https://pharlo.io/docs/): Guides, concepts, and reference.
- [Examples](https://pharlo.io/docs/examples): MCP client configs, REST snippets, webhook signature verifiers, and importable n8n templates.
- [Architecture & security](https://pharlo.io/docs/architecture): How Pharlo is built and how it secures auth, platform tokens, webhooks, and data — for technical evaluators.
- [Roadmap](https://pharlo.io/docs/roadmap): Live vs Planned vs Exploring — only YouTube and Facebook are live and publishable today.
- [FAQ](https://pharlo.io/docs/faq): Concise factual answers — what Pharlo is, live platforms (TikTok is not live), OAuth, billing/credits, retries, webhook signing, hosted/closed-source.
- [Brand assets](https://pharlo.io/docs/brand): Official logos, icons, and Open Graph images for directories and registries (served from https://pharlo.io/brand/).
- [Error catalog](https://pharlo.io/errors.json): Machine-readable error reference for the REST API and MCP server — meanings, HTTP statuses, and the recovery action an agent should take for each.
- [Error handling guide](https://pharlo.io/docs/guides/error-handling): How Pharlo reports errors and how to recover from them.

## Best starting points
- AI agents / MCP clients: connect to the MCP endpoint, then call `list_platforms` and `get_platform_capabilities` to learn supported fields, then `create_assignment` to publish. Fetch the tools catalog for full per-tool docs.
- REST developers: read the OpenAPI spec, authenticate with an API key, and POST to `/api/v1/assignments`.
- Evaluating Pharlo: start with the MCP overview and the landing page.