Back to Ecosystem Pulse
ToolExperimentalMCP
guild
by mathomhaus
Local-first shared memory and coordination for coding agents
Go
Updated Jul 13, 2026
Share:
Summary
Provides shared context, memory, and task coordination for collaborating AI coding agents in a single Go binary. Uses local SQLite and a hybrid keyword + semantic search to surface relevant context and coordinate work across agents. Runs as a local-first MCP-compatible server using the Agent Protocol to keep interactions low-latency and self-hostable, and aligns with the Open Agent Specification (Agent Spec).
Why It Matters
As agent teams coordinate on code, tracing context and past behavior becomes essential to judge reliability and reproduce outcomes. Guild gives a lightweight, local store for agent memories and task state, enabling inspection of interactions and easier reconstruction of agent decisions. That visibility helps teams build early agent-to-agent evaluation and track records without shipping everything to a central cloud service. This also supports cross-agent collaboration via the Agent-to-Agent Protocol (A2A).
Ideal For
Developers building multi-agent coding workflows who want a lightweight, self-hosted context store and coordination layer for MCP-compatible agents. This fits well with established patterns like the Orchestrator-Worker Pattern.
Use Cases
- Coordinating multiple code-specialist agents on a single codebase with shared context
- Persisting agent memories and task state locally for audit and replay
- Enabling hybrid keyword+semantic lookup to provide relevant context to agents during delegation
Works With
claudecodexsqlitemcp
Topics
agent-coordinationagent-memoryai-agentsclaudeclaude-codecodexcursorgolangharness-engineeringlocal-first+3 more
Similar Tools
autogencrewai
Keywords
multi-agent orchestrationagent-memorymcplocal-first