crabtalk/crabtalk

The long-running embedded agent

What it solves

Crabtalk provides a centralized agent daemon that manages the execution of AI agents, the dispatching of tools, and connections to Model Context Protocol (MCP) servers.

How it works

The system consists of a daemon (crabtalkd) that acts as a host. It uses "harnesses" (RV64 ELF binaries) to provide agents with specific, restricted access to resources like the shell, files, and session search. It also supports MCP as a capability for connecting to other programs. Independent binaries called "Apps" connect to the daemon via auto-discovery to create agent-powered experiences.

Who it’s for

Developers who want to build and extend agent-powered services and applications using a structured, daemon-based architecture with restricted tool access.

Highlights

  • Agent daemon architecture
  • Support for Model Context Protocol (MCP) servers
  • Restricted tool access via RV64 ELF harnesses
  • Auto-discovery for standalone agent-powered apps

Related

  • Project
    uvwt/agentdock

    An independent tool runtime that gives AI agents secure, controlled access to execute commands, manage files, and automate browsers across multiple local and remote machines.

  • Project
    apify/mcpc

    mcpc is a lightweight, cross‑platform CLI that gives full access to the Model Context Protocol (MCP) – the standard API for AI agents’ tools, prompts, tasks, and payments. It lets agents and scripts call any MCP operation from a Bash shell, supports persistent sessions, OAuth 2.1 stored in the OS keychain, JSON‑friendly output for piping, dynamic tool discovery via grep, and experimental x402 payment integration.

  • Project
    dbt-labs/dbt-mcp

    An MCP server that gives AI agents the tools to interact with dbt projects, allowing them to execute SQL, analyze data lineage, and manage dbt pipelines.

  • Project
    microsoft/azure-devops-mcp

    An MCP server that gives AI agents structured access to Azure DevOps for managing work items, repositories, and pipelines.

  • Project
    IBM/mcp

    A collection of Model Context Protocol (MCP) servers and tools that enable AI agents to securely interact with IBM products and enterprise resources.