Skip to content
BobSentry

Developers

Integrate runtime authorization in minutes.

Create a sandbox account, register an Agent, issue an Agent-bound runtime credential, call the evaluate API, and inspect signed decision evidence. BobSentry returns the authorization decision; the integration must honor it.

  1. 01

    Create sandbox account

  2. 02

    Register an Agent

  3. 03

    Issue Agent-bound credential

  4. 04

    Send runtime evaluation

  5. 05

    Inspect signed evidence

Get Sandbox API KeyBecome a design partner

Protect your first agent action

Install the SDK and put BobSentry immediately before a consequential action. BobSentry returns the authorization decision; the integration must honor it and perform or withhold the downstream action.

Python

pip install bobsentry

TypeScript

npm install @bobsentry/sdk

View the 10-minute quickstartRuntime integration contract

Integration pattern

Call BobSentry at the tool or action boundary before the consequential operation. BobSentry returns ALLOW, REQUIRE_APPROVAL, or BLOCK. In COOPERATIVE mode, the integration must honor that decision.

  1. 01AI Agent
  2. 02Tool / action boundary
  3. 03BobSentry Runtime Authorization
  4. 04Decision
  5. 05Integration honors decision

Runtime Evaluate API

POST https://runtime.bobsentry.com/v1/runtime/evaluate with normalized action metadata. Deterministic policies return ALLOW, REQUIRE_APPROVAL, or BLOCK. Use an Agent-bound evaluate-only credential.

LIVE API EXAMPLEcurl
curl -X POST https://runtime.bobsentry.com/v1/runtime/evaluate \
  -H "Authorization: Bearer $BOBSENTRY_AGENT_CREDENTIAL" \
  -H "Content-Type: application/json" \
  -d '{
    "action": "deploy_production",
    "metadata": {
      "environment": "production",
      "target": "github-deploy-workflow"
    }
  }'

Integration modes

Customer-facing modes today are COOPERATIVE and OBSERVE. There is no customer-selectable ENFORCED mode.

COOPERATIVE

The integration calls BobSentry before the consequential operation and is responsible for honoring the authorization decision.

OBSERVE

BobSentry evaluates and records what policy would decide, but does not control downstream execution.

Signed authorization evidence is the deliverable, not a side effect.

Every recorded runtime decision produces signed authorization evidence—what was requested, which policy applied, the decision returned, and whether human authorization was required. Evidence records who authorized (when applicable) and when. Execution outcomes are recorded separately from authorization. Sample fields below are illustrative—not production identifiers.

LIVE API EXAMPLESigned decision evidence (sample)
{
  "decision_id": "dec_sandbox_8f3a2c1e",
  "decision": "REQUIRE_APPROVAL",
  "policy_id": "pol_external_comms",
  "policy_version": "2026.04.12",
  "approval_status": "pending",
  "action_hash": "sha256:9c1f…a4e2",
  "correlation_id": "corr_sandbox_41bb",
  "timestamp": "2026-07-23T03:14:22Z",
  "signature": "ed25519:…"
}

Decision evidence fields

  • agent identity
  • requested action (action hash)
  • policy evaluation (policy id / version)
  • decision outcome
  • authorization required (yes / no)
  • authorized by (when applicable)
  • timestamp
  • cryptographic integrity (signature)
  • correlation_id

What BobSentry is not

  • Not an AI orchestration framework
  • Not an observability platform
  • Not a chatbot guardrail
  • Not workflow automation
  • Not a generic AI governance dashboard
  • Not a compliance certification product