Compare · Glean
Firmament vs Glean
Use Glean to search the docs your people already wrote. Use Firmament to capture what your agents learn and create, proven against outcomes and served back to every agent.
Glean and Firmament sit side by side. Glean searches the knowledge your people authored across your apps; Firmament owns the knowledge your agents generate from doing the work: the fixes that worked, the signals, the pages they write and keep current. One serves the past, the other grows the present.
Free for a team of three. 30-day trial of everything.
Side by side
 | Firmament | |
|---|---|---|
| Search across company apps (Slack, Drive, Confluence) | ✓ | Not its job |
| Permission-aware retrieval | ✓ | ✓ |
| Serves knowledge your people already wrote | ✓ | and what agents learn |
| Captures what agents learn from real outcomes | – | ✓ |
| Signals: reinforced when it works, retired when it breaks | – | ✓ |
| Agents write and maintain knowledge pages | – | ✓ |
| "Verified" means it worked in production, not a human re-read a doc | – | ✓ |
| Built for agent writes, with human approval | read-mostly | ✓ |
| Connects to off-the-shelf agents over MCP | via a gateway | ✓ |
| Owns the agent-generated knowledge layer | – | ✓ |
We already have Glean. Why add Firmament?
They cover different knowledge. Glean searches what your people wrote: docs, Slack, Drive, tickets. Firmament owns what your agents learn from doing the work: the fix that worked, the signal that a rule is holding or failing, the pages agents write and keep current. Glean finds the past; Firmament grows the present.
Isn't the knowledge in Glean enough for agents?
It's half of it, and a valuable half. But the knowledge that decides whether an agent succeeds, what actually worked last time and what to never do again, usually isn't written in any doc. It's learned by doing. That is the part Firmament captures, verifies, and serves back.
What does "verified" mean here?
For a search tool, verified means a human re-read a document. For Firmament, verified means it worked in production: reinforced every time it holds, retired the moment it breaks. The knowledge carries a track record, not just an author.
Can agents write to it, or only read?
Both, with governance. Search is built for retrieval over what already exists. Firmament is built for agents to write back what they learn and read it next time, with a human approving anything before it reaches a team.
In fairness
When Glean is the right tool
If what you need is company-wide search so people and agents can find the docs, messages, and tickets your team already wrote, Glean is strong at exactly that. Firmament is for the knowledge that isn't written down yet: what your agents learn from doing the work, captured and proven.
Compare Firmament with