FeedGPT documentation

A dependable data layer for your AI agents

FeedGPT ingests JSON feeds, standardizes their fields, preserves the raw evidence, and serves the resulting records through REST and MCP. Analysis and conversation stay in the AI client you already use.

Open sources

Core workflow

Create a source

Choose a built-in or custom platform. FeedGPT gives the source its own webhook URL.

Send JSON

POST one payload or a bulk array. FeedGPT retains the raw payload and extracts the configured fields.

Query standardized posts

Read the normalized records through REST or connect an external AI agent through MCP.

Maintain the mapping

Inspect extraction gaps, test changes, apply proposals, and reprocess stored raw payloads.

Authentication

Webhook ingestion uses its source URL and optional signature. Other REST requests accept the organization API key in x-api-key or as a Bearer token. Dashboard requests use the active Clerk organization.

curl -H "x-api-key: YOUR_API_KEY" \
  "https://YOUR_HOST/api/posts?timeframe=week"

Core REST endpoints

POST/api/webhook/{platform}/{sourceId}Ingest one payload or a bulk array

GET/api/sourcesList sources

POST/api/sourcesCreate a source

GET/api/postsQuery standardized posts

GET/api/sources/{id}/mappingRead a source mapping

POST/api/sources/{id}/mappingUpdate a source mapping

POST/api/sources/{id}/reprocessReplay the current mapping over stored payloads

GET/api/sources/{id}/reprocess/{runId}Read queued reprocessing progress

POST/api/sources/{id}/reprocess/{runId}/cancelCancel future reprocessing chunks

MCP for external agents

Connect Claude, ChatGPT, or your own MCP client to /api/mcp/mcp. The tools expose sources, standardized posts, mappings, schema inspection, ingestion simulation, and reprocessing. FeedGPT supplies reliable data and operations; the connected client supplies the reasoning experience.