Changelog

Release notes for Fainflow — newest first.

July 12, 2026 — Run a whole flow's lifecycle from your AI

The MCP server grew from "author drafts" to "run the whole thing" — a connected AI client can now build, wire, enable, and observe a flow end to end without opening the dashboard:

  • Native CLI sign-in. Claude Code (and any loopback-OAuth client) connects with plain claude mcp add --transport http fainflow https://mcp.fainflow.com/mcp — Fainflow accepts the client's automatic loopback callback on any port, so there's no fixed port and no API key to copy. See Use Fainflow from your AI (MCP).
  • Go live from the client. create_api_trigger_route, bind_flow_route, and enable_flow route and turn on an API-trigger flow, enforcing the same plan limit and graph validation as the dashboard. create_api_key / list_api_keys / delete_api_key manage the keys that call it.
  • See whether it ran. list_flow_history and get_flow_execution read a flow's runs (values stripped for privacy); delete_flow_history and delete_flow clean up.
  • Per-flow secrets. set_flow_secret / list_flow_secrets / delete_flow_secret manage the vault an AI-authored flow references as {{vault.KEY}}.

Reliability: connections created after encryption-at-rest now decrypt correctly during flow execution, and flows authored via MCP render on the canvas identically to dashboard-built ones.

July 8, 2026 — Bridges, and Fainflow as an MCP server

The big one: Fainflow now speaks MCP in both directions.

  • Connect any MCP server into Fainflow. Bring your own — or a popular one like GitHub or Atlassian (Jira/Confluence) — with five auth methods including OAuth sign-in (Fainflow stores and auto-refreshes the tokens for you). Its tools become usable inside flows and bridges. See Connect an MCP server.
  • Bridges. Compose tools and context from several MCP connections into one set on a visual canvas — raw tools plus context-assembly pipelines you build by hand — and expose it as a single endpoint an AI agent can call. See Bridge.
  • Use Fainflow from your AI. Fainflow now exposes its own MCP server at https://mcp.fainflow.com/mcp (and a per-bridge URL). Connect it to Claude, Claude Code, Cursor, ChatGPT, VS Code, or n8n — sign in with your Fainflow account and the AI can list and build your flows and bridges, and call every tool from your connections. See Use Fainflow from your AI (MCP).

Everything runs as you: the AI only ever sees your own flows, connections, and bridges, and every action is covered by the same audit logging as the dashboard.

Earlier

Fainflow launched with the visual flow editor, Telegram bot & assistant connections, OpenAI / Claude / Gemini actions, the API Trigger, the encrypted per-flow vault, and billing. Those aren't listed individually here — this changelog starts with the entry above.