Agent Identity
Every agent has a governed identity, a lifecycle state, and a human accountability anchor. The human who sponsors an agent is recorded in every gate decision. That sponsorship cannot be disclaimed.
Govern AI agents before they act.
CLAIIM gives every AI agent a governed identity, binds it to a human accountability anchor, checks every action against policy, and records every ALLOW or DENY in Chron.
AI agents are acting. Most orgs cannot answer:
- Which agent acted?
- Who authorized it?
- Was it allowed by policy?
- What skill governed the action?
- What proof exists afterward?
CLAIIM makes every one of those questions answerable before the action and after it.
Three parts. One control plane.
Policy Gate
Every agent action passes through the gate before it happens. The gate checks identity, skill scope, and active policies. It returns ALLOW or DENY with a reason. Nothing bypasses it.
Chron Proof
Every gate decision -- ALLOW and DENY -- is written to Chron immediately. Chron is designed as an append-only audit trail. The record exists before the action completes.
From ungoverned to accountable.
Without CLAIIM
ungoverned path
Agentdeploy-bot
direct API call
External Serviceproduction API
action executes
Log (after the fact)audit log
xNo proof an action was authorized
xEvery agent can act if it has credentials
xLogs appear after damage is done
With CLAIIM
governed path
Agentdeploy-bot
gate request
CLAIIM Gatepolicy check
Identity
Skill
Policy
ALLOW / DENY
External Serviceproduction API (ALLOW only)
proof recorded
Chronimmutable proof
outcomeALLOWanchors.kondepudipolicystaging-v3
+Human anchor on every decision
+Policy checked before execution
+Proof exists before action completes
Gate your first agent in 10 minutes.
1
Provision agentRegister the agent in CLAIIM. Assign it to an organizational boundary and bind it to a human accountability anchor.
2
Define skill and policyCreate a versioned skill: the email identity, system prompt, allowed actions, and forbidden actions. Set org-wide policies for what agents may never do.
3
Gate every actionYour agent calls the gate before acting. CLAIIM checks identity, skill scope, and active policies. The agent receives ALLOW or DENY with a reason. The gate is fail-closed.
4
Review Chron proofEvery decision is recorded with agent, human anchor, organizational boundary path, skill version, and policy version. Auditors and regulators see the full chain.
Built for accountability, not just visibility.
Human accountability anchor
Every agent is sponsored by a named human. Their identity travels with every gate decision. Regulators see both the agent and the person who authorized it.
Organizational boundaries
Agents belong to an organizational boundary called an AIORG. Agents cannot cross boundaries unless explicitly granted. Hierarchy is enforced at the gate.
Versioned skills and policies
Every change to a skill or policy creates a new version. Agents hold a pinned version -- silent updates are not possible. Chron records which version governed each decision.
Two-person control for privileged changes
Sensitive management operations require a second admin to approve. All actions inside the elevated window are tagged in Chron with the session ID.
Deployment
Runs in your environment.
CLAIIM deploys inside your infrastructure and uses your database. Chron, policies, agent records, and gate decisions remain under your control. Nivaya does not receive agent activity data.
Read install guideYour infraDocker Compose or Kubernetes
Your databasePostgreSQL you control
No data leavesNivaya receives no agent activity
Evaluation Preview now. Professional and Sovereign coming.
Evaluation Preview
Available now- Gate enforcement and Chron
- Organizational hierarchy
- Versioned Skills and Policies
- Two-person control
- Admin UI
- Docker Compose install
- Python SDK
Single nodePostgreSQL onlyEmail and password auth
Professional
In development- Everything in Evaluation Preview
- Kubernetes Helm chart
- Active-active HA
- SAML and OIDC federation
- MSSQL support
- Cross-org grants UI
Sovereign
In development- Everything in Professional
- Air-gap operation
- mTLS agent mesh
- Signed USB update bundles
- EU AI Act evidence pack
- ISO 42001 evidence pack
- DORA incident trail
Ready to gate your first agent?