What it is
Cognition Mesh is the governance layer over the AI you bring. Not a model. Not a proxy between you and your vendor. A control plane that decides, before an AI acts, whether that action is allowed, escalated to a person, or refused, and keeps a tamper-evident record either way.
You bring your own models and your own contracts. We govern what those AIs are allowed to do, what they're allowed to know, and what evidence proves both.
What it does, and refuses to do
It does
- Checks every consequential action against policy before it runs.
- Routes decisions that need a person to a named, accountable person, through the approval channels you already use.
- Produces a signed, recomputable receipt for every decision.
- Governs every AI you run, across their whole lifecycle, not one tool call at a time.
- Runs on infrastructure you control.
It refuses
- To let an AI take a consequential action nobody authorized.
- To let AIs authorize other AIs with no human in the chain.
- To give you an editable log and call it evidence.
- To lock you to one vendor or one model.
- To sit in the cloud with your data when you'd rather it didn't.
Where it runs
On your infrastructure. On-prem, your VPC, your laptop. Every competitor we've found is cloud-only; we made the opposite choice on purpose, because the people who need this most are the people who can't send their AI activity somewhere else.
Receipts
Every decision produces a record you can recompute and hand to a regulator. Not a log a developer can edit. The EU AI Act's tamper-evident logging requirement has been in force since August 2026; this is what that looks like in practice.
Tiers
Solo
One person, one machine. Docker install, trial to start.
Pro
A team and its repos. Adds the shared bus, dispatch, and management console.
Enterprise
The whole organization. Multi-tenant, fleet-wide enforcement, your approval chain up to the board.
Pricing and the full comparison live at cognition-mesh.com.
Contact
Talk to a person. fred@cognition-mesh.com