Busabase
The open-source database and workspace for your AI agents
About Busabase
Busabase addresses a problem that grows quietly as teams adopt agents: the output ends up trapped in chat histories, owned by whoever ran the prompt and invisible to everyone else. Busabase gives agents a database and a workspace instead, so what they produce lands as structured records in shared spaces rather than as messages in a transcript. People and agents work in the same place, agents you already use plug in and keep what they generate, and a change-request and review queue puts a human between agent output and accepted fact.
The engine is MIT-licensed and free forever whether you run it yourself or Busabase runs it for you, which means the core is genuinely open rather than open-core marketing with the useful parts withheld. Self-hosting carries no seat limit at all. The paid product is Busabase Cloud, which sells hosting, shared spaces and the review queue on a flat per-seat basis with no metered usage, so costs are predictable in a way that usage-billed agent infrastructure usually is not. For teams wary of lock-in, the combination of an MIT core and a hosted option with a clean exit is the main argument in its favour.
You create spaces, which are the structured containers agent work lands in. Busabase describes eight kinds of work you can start keeping immediately, so the model is not a blank database you must design before getting value from it.
You then connect the agents you already use. Rather than replacing your agent stack, Busabase sits underneath it as the place output is written, so results become records with fields rather than paragraphs in a thread. People work in the same spaces, which is what makes the output reviewable by someone other than the person who prompted it.
Changes pass through a review queue. Agent output arrives as change requests that a human accepts or rejects, with the full queue on paid tiers and multi-stage review on Pro, so there is a controlled boundary between what an agent proposed and what the team treats as true. You can run all of this on your own infrastructure under the MIT licence, or on Busabase Cloud with hosting and shared spaces managed for you.
- •Structured records instead of chat logs - agent output is stored as data in spaces rather than messages, which is what makes it searchable, reviewable and reusable later.
- •MIT-licensed engine - free forever and self-hostable with no seat limit, so the core product carries no lock-in and no per-user cost if you run it yourself.
- •Human review queue - change requests gate what agents produce, with full review on Plus and multi-stage review on Pro, keeping a person between proposal and accepted record.
- •Shared spaces for people and agents - one workspace both humans and agents write into, so agent work is visible to the team rather than private to its operator.
- •Bring your existing agents - plugs in underneath the agents you already run rather than asking you to adopt a new agent framework.
- •Flat per-seat cloud pricing - no metered usage on Busabase Cloud, so the bill does not move with how much your agents happen to do in a given month.
Busabase suits engineering and operations teams running several agents whose output currently lives in scattered chat histories, organisations that need a human approval step before agent output is trusted, and teams with data residency or lock-in concerns who want the option to self-host. The MIT licence makes it a reasonable choice for anyone who has been burned by a hosted tool changing its terms, and flat per-seat pricing suits teams whose agent volume is unpredictable.
It will not suit everyone. The idea of a system of record for agents is a new and unproven category, and it is entirely possible that a team's existing database, ticket tracker or wiki already does enough. Small teams running a single agent will find the review queue and space model more structure than the problem deserves. Self-hosting means operating a database, so the free path has real running costs in time. Organisations wanting SSO, SAML, audit log export or a data processing agreement are pushed into an unpriced enterprise tier that requires contacting sales. And the product is new, with no reviews yet, so the review workflow in particular is worth trialling on real work before it becomes load-bearing.
Pricing
- Free$0/mo
- Plus$12/mo
- Pro$24/mo
- EnterpriseContact sales
From the vendor pricing page, 2026-10-15
