@alfe.ai/mcp-server
v0.2.12
Published
Local stdio MCP server bundled with the Alfe CLI. Auto-registered as `alfe-platform` in the bundler store; surfaces Alfe platform tools to the agent runtime via @alfe.ai/mcp-tools using AgentApiClient against public /agent/... endpoints.
Readme
@alfe.ai/mcp-server
Local stdio MCP server bundled with the Alfe CLI. Auto-registered as alfe-platform in the bundler store; surfaces Alfe platform tools to the agent runtime via @alfe.ai/mcp-tools using AgentApiClient against public /agent/... endpoints.
Part of Alfe — the operating system for AI agents: build, deploy, and run agents with persistent memory, identity, integrations, and channels. See the documentation to get started.
Install
npm install @alfe.ai/mcp-serverProfiles
The executable defaults to the OpenClaw-safe integrations-only surface:
alfe-mcp-serverClaude Code has no Alfe plugins, so its explicit profile adds memory, voice, and messaging tools:
alfe-mcp-server --profile claude-codeThe default profile must remain integrations-only. OpenClaw already receives the other capabilities from its native Alfe plugins, and exposing them here as well would duplicate its tool surface.
The server reads credentials through @alfe.ai/config. Library callers may
inject a pre-bound AgentApiClient, its API URL, and an identity into
createServer for tests or alternate hosts. The main client and URL must be
provided together. A distinct voiceApiUrl also requires a voiceClient; the
package never reads the local API key and attaches it to an arbitrary injected
destination.
main() attaches a transport and returns the connected server without changing
process state. The packaged executable owns fatal logging, exit codes, and
SIGINT/SIGTERM shutdown.
Links
- 🌐 Website: https://alfe.ai
- 📚 Docs: https://docs.alfe.ai
