Glen MCP server
Shared memory for a team's coding agents: recorded sessions, Slack, GitHub and tickets, recalled with a source on every claim.
Glen is a hosted memory layer shared by everyone in an organization. Plugins for Claude Code, Codex, Cursor and other agents record every session through hooks, and integrations pull in public Slack channels, GitHub pull requests and reviews, Linear issues, chosen Notion pages and call transcripts from Fireflies, Fathom and Granola. A background pipeline turns that material into observations: individual facts and decisions, each tied to who said it, where and when. Access control is applied per observation at recall, so an agent only receives what its user can already see, and when a newer decision contradicts an older one the older one stops surfacing.
The MCP server at https://mcp.tryglen.com/mcp speaks Streamable HTTP and authenticates with OAuth 2.1 or a personal API key. glen_search returns one synthesized answer with a citation on every claim rather than a list of raw chunks, and glen_code_search walks from a line range to the conversations that wrote it, using git blame and a Glen-Conversation commit trailer. The same server holds a skills library, an artifact library, and the controls for forgetting evidence and going off the record. The glen CLI reaches everything the MCP server does. Glen is a managed service with no self-hosted build; the first 15,000 memories are free.
Quick install
Copy-paste configs are provided for all 8 supported clients. Pick your client below.
Add to ~/.claude.json
{
"mcpServers": {
"glen": {
"type": "http",
"url": "https://mcp.tryglen.com/mcp"
}
}
}claude mcp add --transport http glen https://mcp.tryglen.com/mcpHeads up
- First tool call opens a browser to authorize.
Available tools
| Tool | Description |
|---|---|
| glen_search | Search the organization's shared memory and return one synthesized answer with a citation on every claim; each citation reference can be passed to glen_forget. |
| glen_code_search | Trace specific lines of code back to the agent conversations that produced them, from the commit SHAs git blame reports and any Glen-Conversation commit trailers. |
| glen_forget | Exclude verified wrong, stale or irrelevant evidence from future recall; the raw messages stay in transcript history. |
| glen_restore | Restore previously forgotten evidence so it is eligible for recall again. |
| glen_import | Import one turn of a past coding-agent conversation into the organization's memory. |
| glen_skill_search | Find a reusable skill in the organization's library by intent or author. |
| glen_skill_use | Fetch a skill's instructions by its exact slug, for the agent to follow. |
| glen_skill_create | Save a new skill (a markdown prompt) to the library, private by default or shared with the organization. |
| glen_skill_share | Share a skill you authored with teammates by email or with the whole organization. |
| glen_skill_delete | Permanently delete a skill you authored. |
| glen_dependency_search | Search the organization's existing skill dependencies (tools and runtimes) so a new skill reuses a match. |
| glen_artifact_search | Find a saved document in the artifact library by intent or keywords, among artifacts you authored or that were shared with you. |
| glen_artifact_get | Fetch a saved artifact by its exact slug. |
| glen_artifact_create | Save a Markdown or HTML document to the artifact library, private by default. |
| glen_artifact_share | Share an artifact you authored with teammates by email or with the whole organization. |
| glen_artifact_delete | Permanently delete an artifact you authored. |
| glen_status | Show the user's Glen mode (on, off, incognito or silent) and whether skill suggestions are on. |
| glen_set_recording | Set the user's mode account-wide: on, off, incognito (recall but record nothing) or silent (record but surface nothing). |
| glen_set_suggestions | Turn skill-creation and skill-use suggestions on or off for the user. |
| glen_list_orgs | List the organizations the user belongs to and mark the one this connection reads and writes. |
| glen_invite | Invite teammates by email to the active organization so they share its memory. |
| glen_feedback | Send a bug report or product feedback about Glen to the Glen team. |
What you can do with it
Give every agent on a team the same history
A decision one developer reached with Claude Code in March reaches another developer's Codex session in June, with the conversation it came from cited, without anyone writing it into AGENTS.md.
Find out why a block of code exists
glen_code_search takes the commits git blame reports for a line range and returns the agent sessions that produced them, ranked against your question, so the agent reads the reasoning instead of guessing from the diff.
Start a session with what the team already knows
In Claude Code and Codex, Glen reads the live session and injects relevant context before the agent starts searching the repository, which is where most of the token savings in Glen's 100-PR replay came from.
FAQ
- Is it free?
- The first 15,000 memories are free with no credit card. Paid plans are per seat.
- Does it support remote/OAuth?
- Yes. It is a remote server at https://mcp.tryglen.com/mcp over Streamable HTTP, with OAuth 2.1 discovery for clients that support it and a personal API key for those that don't.
- How is it different from Mem0 or Supermemory?
- Those are memory APIs and servers that store what an agent or app chooses to write, usually scoped to one user. Glen records every session and connected tool automatically, scopes memory to the whole organization with per-observation access control, reconciles contradicting decisions, and returns a cited answer instead of raw matches.