Named agents
Every AI agent is registered with a governed identity. It is assigned to a team or department and bound to a named person who is on record for every decision the agent makes.
The product
Three things that make AI agents accountable.
Registered agents have a governed identity. Requests sent through the gate are checked against policy. Chron keeps an append-only application audit record in your environment. Independent verification is in development.
CLAIIM Gate
agentsupport-bot
ownerr.okafor · support lead
teamcustomer support
ALLOWsend customer reply
DENYissue refundneeds human approval
Chronappend-only proof
09:12:44ZALLOWsend customer reply
09:12:51ZDENYissue refund
owner: r.okafor · policy: support-policy-v2 · team: customer support
Three parts. Working together.
Policy before action
Requests submitted through the gate are checked against your rules. CLAIIM returns ALLOW, DENY, or HOLD with a reason. Enforcing the decision requires a controlled execution path.
A record of every decision
Gate decisions are recorded in your audit log. The record lives in your environment and is append-only in the application. Independent verification of the record is in development.
Ungoverned versus accountable.
The same agent. One path has a governance layer. One does not.
Without CLAIIMungoverned path
Agentdeploy-bot
direct API call
External serviceproduction API
action executes
Audit log (after the fact)no proof of authorization
xNo proof an action was authorized
xAny agent with credentials can act
xLogs appear after damage is done
With CLAIIMillustrative gate path
Agentdeploy-bot
gate request
CLAIIM Gateidentity + skill + policy
ALLOWdeploy:stagingDENYdeploy:production
proof recorded first
Chronappend-only proof
anchors.kondepudipolicyprod-deploy-v3
+Human anchor on every decision
+Policy checked before execution
+Decision recorded for review
What happens when an agent acts.
Integrated agents submit requests through this sequence. Enforcement depends on the execution path and who holds the destination credentials. See integration coverage and product status.
01
The agent sends a request
When an agent wants to take an action, it sends a request to CLAIIM before touching any system. The request includes the agent identity, the intended action, and the scope it is operating in.
02
CLAIIM checks identity and scope
CLAIIM confirms the agent is registered, that it belongs to the team or department it claims, and that it is operating inside its assigned boundary.
03
The action is matched against policy
The action is evaluated against the active policy set for this agent. CLAIIM checks whether it is allowed, blocked, or held for a person to approve.
04
The gate returns a decision
ALLOW authorizes the evaluated request, DENY refuses it, and HOLD requires review. A gate-only decision does not prove that execution happened or was prevented elsewhere.
05
Chron records the decision
Chron records the decision, agent identity, responsible person, rule version, and timestamp. Gateway receipts describe reported delivery; they are not independent verification of downstream success.
Every Chron entry contains:
agent
Which agent sent the request and what it claimed to be doing.
owner
The named person in your team who is on record for this agent.
decision
The gate decision and reason. A decision does not independently verify downstream execution.
policy version
The exact rule version in effect at the time. If rules change later, the record shows what applied.
Runs in your infrastructure. Maps to how your teams work.
The preview stores governance records in your own deployment. You control database, network, integration, and support access. Tenant isolation conversion is in progress and has not completed acceptance certification.
What runs in your environment
The gate
Receives integrated action requests and returns a policy decision. The controlled HTTP gateway executes authorized requests using its destination credential.
Policy store
Holds the rules for each agent and scope. Every change creates a new version. Policies are versioned so silent edits are not possible.
Chron
The append-only audit log. Every gate decision is written here. Records stay in your own database under your control.
Identity registry
Holds registered agents, their team and department assignments, and the responsible person bound to each agent.
How it maps to your teams
Teams and departments are the boundary
Agents are assigned to a team or department for policy evaluation. These scopes do not contain local shell commands, file edits, or actions using separate destination credentials.
A named person is on record for every agent
Each agent is sponsored by someone in your team. Their name travels with every gate decision. Accountability is always traceable to a real person.
Policies belong to the team that owns the agent
Rules are set at the team or department level and apply only to agents in that scope. A change to finance policies does not affect engineering agents.
Peer-approved elevation for protected operations
Supported privileged operations require an elevated session approved by another person. Organisation proposals have a separate peer-confirmation flow. Session approval is not a second approval of every individual change.
Ready to govern your first agent?
Preview is free and runs in your own environment.