Docs MCP — live today
Lets agents query the Bavlio documentation by tool calls (search, fetch page). Auto-generated from this site.
API-action MCP — roadmap
Lets agents take actions against the live Bavlio API (create campaigns, send email, search leads). Planned.
Docs MCP — live
Auto-published by the docs platform at:Quick install (Cursor)
Add to your Cursor MCP config:Quick install (Claude Desktop)
Open Claude Desktop’s MCP config and add:What this covers vs. doesn’t
Covers: every page on this site — Quickstart, Authentication, Errors, Pagination, Rate Limits, Webhooks, the full API Reference (190 endpoints with request/response schemas), and the OpenAPI spec. Doesn’t cover: taking actions against the API. Agents using only the docs MCP can read docs but can’t, for example, create a campaign or send an email. For that you need the API-action MCP below — or just call the REST API directly with abav_live_ key.
API-action MCP — roadmap
A separate MCP server that wraps the Bavlio REST API as agent-callable tools. Lets an agent create campaigns, send emails, search leads, etc. without writing HTTP plumbing.Planned install path
bav_live_ API key.
Planned tool surface
Initial tool set targets the most common agent workflows: campaign lifecycle, lead management, send + reply, plus the two stateless email primitives.Design notes
- Untrusted-content envelope. Tools that surface third-party content (inbound replies) wrap responses in
{ __untrusted_third_party_content: true, content: ... }. The fence is a machine-readable signal that a trust boundary was crossed. Same pattern as@bavimail/mcp-server. - No client-side caching of API keys. The server reads the API key on every call; key rotation takes effect without restart.
- Tool annotations. Every tool ships with
readOnlyHint,destructiveHint, andidempotentHintper the MCP spec.