Lock out risky agent actions. Prove every decision.
Put agents into production and prove every consequential action was approved by a person. Routine work runs. Risky actions wait for a verified human. Every decision is signed and verifiable offline.
Free posture scan · No credit card · Maps what your agents can reach
-
notion.pages.readAllowed
-
github.pulls.mergeReleased
-
stripe.refunds.createHeld
- Agent
- support-bot
- Action
- stripe.refunds.create
- Amount
- $4,800.00
- Policy
- Refunds over $500 need a human
- Requested
- 14:02:11 UTC
Try it: approve or deny the refund.
Monitoring shows what your agents did. Interlok proves who allowed it.
Agents now hold real keys: payments, production, cloud permissions, customer data. Interlok gates actions, not agents. Only the few you can't take back stop for a person.
Reads, search, summaries, drafts
Flows straight through.
Routine writes
Flows through and is signed into the ledger.
Payments, production writes, deletes, permissions, model promotions
Stops and waits for a human decision.
Actions no one should approve
Blocked by policy, and recorded.
One gate and one signed record, whatever you answer to.
Software and AI teams
Agents acting in production, and security reviews that ask what they did.
Defense suppliers
Agents and pipelines near CUI, and an 800-171 affirmation that needs proof. Defense →
Regulated industries
Finance, health and life sciences, where a named approver is the rule.
Plants and infrastructure
Firmware and grid changes, where a wrong change is physical.
Find. Gate. Decide. Prove.
-
01
Find
Scan your agents' MCP servers, tools and credentials. Every ungated path to money, data and production is flagged.
interlok scan --ci -
02
Gate
Route tool calls, deploys and model promotions through policy you keep as code. Start in shadow mode, then enforce.
p95 25 ms on an allow -
03
Decide
A held action waits for a person, in Slack or the console, verified by your IdP.
an agent never approves an agent -
04
Prove
Every decision is signed and hash-chained with keys you hold. Every model run gets a receipt too: which model, which provider, which region, with prompts hashed and never stored.
signed receipt · offline verify
Works with the agents you already build with: Copilot Studio, Agentforce, OpenAI, Claude, Google ADK, Bedrock AgentCore, LangGraph, CrewAI, Copilot and Cursor. See how each connects →
- Customer-held keysYour keys sign every record.
- Offline verifierCheck evidence without us.
- Hourly public anchorsA Merkle root anyone can check.
- Inference receiptsBuilt from your telemetry, never in the inference path.
- Compliance packsSOC 2, ISO 42001, NIST SP 800-171, EU AI Act.
Scan free. Prove with Evidence Connect. Gate when you are ready.
Posture scan
Free- MCP server and tool discovery
- Ungated-write-path detection
- Offline verifier for any Interlok bundle
Evidence Connect
Keep your gateway- Signed evidence from the gateway you already run
- Inference receipts: model, provider and region for every run
- Bundles mapped to SOC 2, ISO 42001, NIST SP 800-171 or the EU AI Act
Gate
Production enforcement- Everything in Evidence Connect
- Holds for an IdP-verified person, shadow mode first
- Model promotions held for a person, like a risky deploy
See what your agents can reach.
Run the free scan and get a map of your agents and tools, with ungated write paths flagged.
Free · No credit card