Model Context Protocol
What MCP gives you
MCP is an open standard for connecting AI agents to external tools and data. An MCP server exposes a set of tools; an MCP client calls them. Flowly is both.
Point Flowly at any MCP server β local (npx / uvx) or remote (HTTP / SSE, with OAuth) β and the agent calls its tools like any built-in.
flowly mcp serve exposes your conversation history to any MCP client over stdio β Claude Desktop, Cursor, Claude Code, another agent.
OSV malware checks, filtered subprocess environments, credential redaction, a circuit breaker, and OS-level sandboxing wrap every server.
A curated catalog of version-pinned servers (Context7, GitHub, Linear, Playwright, β¦) you install with a single command.
Quick start
Browse the catalog, install a server, and start using it:
flowly mcp catalog # browse curated, ready-to-install servers
flowly mcp install context7 # install one (it probes the connection)
flowly mcp list # see what's configured
flowly mcp test context7 # connect + list its toolsStart a new session and the agent has the server's tools β for Context7, that's mcp_context7_resolve_library_id and mcp_context7_query_docs. In the terminal chat, run /mcp for the same thing with arrow keys.
flowly restart) or start a new flowly agent / TUI session. The /mcp modal restarts the gateway for you.Two directions
Flowly as a client β the common case. You add MCP servers and the agent uses their tools. Connecting servers β
Flowly as a server β the other direction. flowly mcp serve turns Flowly into an MCP server so other agents can read your conversation history and (with --allow-writes) send messages on your behalf. Flowly as an MCP server β