SendRaven

• CollectionCommunication
• Classificationagent-native
• MCPAvailable
• Agent SkillUnavailable
• InterfacesMCP · REST · webhook · openapi · STDIO · http · API

• Website ↗
• Docs ↗
• Official repo ↗
• Full dossier on GitHub ↗

What it does

SendRaven is an email platform whose stated primary caller is an AI agent. An agent sends transactional email from the customer’s own verified domain, receives the replies, and reads them as thread state: each reply is joined to its conversation on Message-ID (never on subject), quoted history is stripped, and a thread carries an awaiting_reply flag, so the agent can ask “what needs an answer?” and reply in the same thread.

The second half is the constraint model. Limits live on the credential the agent holds, not on the workspace: a per-key daily send cap, a recipient allowlist, and an approval hold that turns every send from that key into a draft a person releases. A key cannot edit its own limits, mint a looser key, or approve its own drafts.

The same platform also runs campaigns, automations and ordinary transactional mail (it replaced the email providers of its builders’ own products), so it is not agent-only; agents are the primary consumer on the homepage, the docs and the MCP page.

Agent-facing interfaces

• Interaction pattern: MCP server (remote Streamable HTTP with OAuth, or stdio with an API key); REST for everything else
• Catalog protocols: MCP, REST, webhook, openapi, STDIO, http, API
• MCP: ✅ Available
• Agent Skill: ⚠️ Not yet published
• Onboarding type: mcp

How agents onboard

Interaction pattern: MCP server (remote Streamable HTTP with OAuth, or stdio with an API key); REST for everything else

npx -y @sendraven/mcp   # with SENDRAVEN_API_KEY=sk_live_... in the environment

See the full onboarding section in the dossier.

Protocol surface