Glen pricing, free plan and limits

Glen is a shared memory layer for AI agent teams, a Y Combinator Summer 2026 company. It recalls team context from Slack, GitHub and past agent sessions into each prompt. Sign-up is open and free to start with no credit card, and the vendor's own replay test reports 29% fewer tokens.

Glen, built by Glen Labs, gives coding agents a shared organizational memory. It installs with one npx command, pulls in Slack and GitHub history, and feeds relevant prior work into each prompt. The vendor's August 2026 replay of 100 pull requests reported tasks finishing 21% faster at parity quality, and the free tier needs no credit card.

Maker: Glen · Protocol: MCP · Auth: oauth

Compatible agents: Claude, Claude Code, Codex, Cursor, Pi, any MCP client

Required runtime: Any MCP-compatible agent host (Claude Code, Cursor, Codex, or custom framework)

About Glen

Glen is a shared memory skill for AI agents from Glen Labs, founded in 2026 by Nikos Dritsakos and backed by Y Combinator's Summer 2026 batch. It began as a Model Context Protocol server with two tools, recall and store, and today the company describes it as a way to surface organizational context into every prompt so that agents work better, faster and on fewer tokens.

Setup is one command, npx @tryglen/cli@latest install, or a setup prompt pasted into the agent you already use: the vendor's site lists Claude Code, Codex (OpenAI Codex), Cursor and Pi, plus a prompt for custom agents. From there Glen reads public Slack channels, shared team transcripts and GitHub activity, then injects the prior work that bears on a task before the agent starts. A recalled item shows up labelled "Glen recalled", so you can see which thread or session shaped the answer. The same store powers a PR reviewer that queries a clone of the agent that wrote the change, a way to pick up a teammate's session, and shared team skills that any agent can load by name. Composio is one of the teams the site names as a user; see its own Composio skill page for what it sells.

The one published measurement comes from Glen itself. In an August 2026 post the company reverted the last 100 merged pull requests in a workspace that had run Glen for two months, then had Codex 5.5 re-implement each one with and without team memory, five runs per arm. It reports 29% fewer tokens and 21% faster wall-clock time, with quality at parity or better. That is a vendor test on its own workspace, not an independent benchmark, so treat the size of the gain as a claim to reproduce on your codebase.

On access and cost, the homepage now offers a free start with no credit card, which replaces the earlier waitlist-only beta. Glen has not published any paid prices. Per-memory access control applies at recall, DMs and private channels are never read, and an incognito mode keeps a session out of the store. On training, the policy says Glen does not use what you store for its own models, though content does go to AI and infrastructure subprocessors.

Who should look elsewhere: teams that need open-source code or self-hosting will not find it here, because Glen's code is proprietary. A solo developer who only wants memory of their own sessions is better served by per-user memory such as Mem0, and anyone wanting a full agent runtime with tiered memory should compare Letta. Teams whose real knowledge lives in private channels and DMs will also get less from Glen by design. To choose the agent that sits on top of it, read our comparison of Cursor and Claude Code, the explainer on what Cursor is for, and the profiles of Windsurf and GitHub Copilot. For the wider landscape, see our guide to the best agentic AI tools of 2026.

Key Features

• Org-wide shared memory: Every agent and teammate in an organization reads from the same store, so recalled context from a prior Slack thread or agent session reaches the next task before it starts.
• One-command install: The CLI installs with npx @tryglen/cli@latest install, and a pasted setup prompt configures Claude Code, Codex, Cursor, Pi or a custom agent.
• Recall labelled in the answer: Recalled items appear as "Glen recalled" with the thread or session behind them, so a reader can trace which prior work shaped a suggestion.
• Per-memory access control: Each memory keeps its source's permissions, which limits an agent to what its user is already cleared to see.
• Public-sources-only capture: Capture is limited to public Slack channels and shared team transcripts; direct messages and private channels are never read, according to the vendor.
• Incognito mode: One click from the menu bar keeps a session out of the memory store entirely.
• PR reviewer built on the author agent: The pull-request reviewer and code search query a clone of the agent that made the change, which the vendor says returns the reason for a change and not only the diff.
• Shared sessions and team skills: Teammates can pick up each other's sessions and load a colleague's skill by name, for example glen use skill code-like-jason.

Use Cases

• Debugging with what the team already worked out: An agent asked why a nightly eval job times out is shown the Slack thread that cut the shards and the session that opened the related issue, instead of re-reading the repository from scratch.
• Reviewing a change against its history: The PR reviewer asks the agent that wrote a change why a retry delay moved from 5 to 30 seconds and gets the incident that prompted it.
• Picking up a teammate's work: A new engineer's agent loads a colleague's session or named skill, so onboarding starts from how the best person on the team actually works.
• Following the latest decision: When a launch date moves through three Slack updates, the agent follows the decision that stuck rather than the first one it finds.

Install

npx @tryglen/cli@latest install

Requirements

• A free Glen account created at app.tryglen.com (the first 15,000 memories are free, no credit card)
• A coding agent to attach it to, from the list on the vendor's setup section
• Slack and GitHub connections if you want team history captured, limited to public channels and shared transcripts

Actions

Recall

Retrieves the org observations most relevant to the current agent task from the shared knowledge pool.

{
  "method": "tools/call",
  "params": {
    "name": "recall",
    "arguments": {
      "context": "Reviewing a pull request for the billing service"
    }
  }
}

• context (string) — required: A plain-English description of the current task or situation. Glen uses this to retrieve the most relevant stored observations for this turn.

Store

Writes a new observation from the current agent turn into the shared org knowledge pool for future agents to recall.

{ "observation": "Stripe calls from the billing service need an X-Idempotency-Key header", "context": "Billing code review" }

• observation (string) — required: The fact, convention, or decision the agent learned this turn, written as a self-contained statement a future agent can read without extra context.
• context (string): Optional description of the situation in which this observation was made, used for future relevance scoring.

How to Invoke

Agents add Glen as an MCP server in their host configuration. Each turn, the agent calls the recall MCP tool with the current context to retrieve relevant org observations, and the store tool to write new learnings back. Both run in a single round trip per turn.

Pricing

Glen's homepage (read 26 October 2026) offers "Start free" with the first 15,000 memories free and no credit card required. The privacy policy mentions paid plans, but no prices, plan names or per-memory rates are published, so the cost beyond the free allowance is unknown. Earlier HokAI notes described a closed beta with a waitlist; the homepage no longer says that. Compare with [Mem0](/hub/tools/mem0) and [Letta](/hub/tools/letta), which publish their own plans.

Strengths

• Context from Slack, GitHub and earlier agent sessions reaches each prompt without a documentation step.
• One npx command or pasted prompt installs it for Claude Code, Codex, Cursor and Pi.
• Permissions follow each memory's source, and DMs and private channels are never read.
• A free tier with no credit card lets a team test it before paying.

Weaknesses

• The privacy policy refers to paid plans but lists no prices, which leaves the long-run cost unknown.
• The speed and token gains come from the vendor's own replay on its own workspace, with no independent test.
• Proprietary code with no self-hosted or open-source option, so teams cannot audit the memory layer.
• Early-stage product founded in 2026 with no long-term support track record.

View the official Glen skill page