Korg Recall Mcp
Semantically search and recall context across Claude Code, Codex, and Grok sessions so one agent session remembers what others decided.
Overview
korg-recall-mcp is an MCP server for the Build phase that enables semantic memory recall across multiple AI coding agent vendors.
What is this MCP server?
- NPM @korgg/recall-mcp v0.1.1 stdio MCP server from New1Direction/korg
- Semantic recall across Claude Code, Codex, and Grok—not exact string match on filenames
- Complements korg-ledger audit trails with meaning-based retrieval
- Part of Korg cross-vendor cognition stack alongside introspect wrappers
- Reduces re-explaining architecture every time you switch coding agents
- Server version 0.1.1
- NPM identifier @korgg/recall-mcp
- Documented cross-vendor scope: Claude Code, Codex, and Grok sessions
Community signal: 3 GitHub stars.
What problem does it solve?
You switch between Claude Code, Codex, and Grok and each session starts cold even though the answers already exist in another tool's history.
Who is it for?
Multi-agent solo builders who want a single semantic memory layer over disparate coding assistants.
Skip if: Teams on one vendor only with adequate built-in history, or orgs that cannot store session-derived embeddings locally.
What do I get? / Deliverables
After install, agents query prior work by meaning and pull the right snippets into the active session without manual transcript archaeology.
- Semantic recall queries usable from MCP-enabled agents
- Reduced context re-entry when switching between coding assistants
- Bridge layer in a multi-tool Korg cognition stack
Recommended MCP Servers
Journey fit
Cross-vendor memory lands in Build agent-tooling because you install it where agents are extended, even though recall pays off later in Operate. Agent-tooling covers MCP memory layers that persist cognition outside a single chat vendor.
How it compares
Cross-vendor semantic memory MCP, not a vector DB admin console or a single-repo RAG skill.
Common Questions / FAQ
Who is korg-recall-mcp for?
Solo developers juggling Claude Code, Codex, and Grok who need shared recall instead of siloed chat logs.
When should I use korg-recall-mcp?
Use it when you repeatedly re-explain project context after changing agents or when iterating on bugs discussed under different wording.
How do I add korg-recall-mcp to my agent?
Install @korgg/recall-mcp from npm, add io.github.New1Direction/korg-recall-mcp as a stdio MCP server in your client config, and connect it alongside your other Korg tools if needed.