HOW IT WORKS

How obsvr governs agents and model calls

Decide before dispatch. Preserve the signed decision that followed.

THE PIPELINE

Four stages. Explicit boundaries.

Intercept

The Obsvr SDK intercepts documented provider, framework, and tool routes. Explicit wrappers provide the clearest boundary; supported zero-code hooks require the documented runtime and load order.

Enforce Policy

The policy engine evaluates the call against your rules before dispatch. On enforcing routes, a block means the governed provider or tool is not invoked. Deterministic. No model judging another model.

Sign & Persist

Each emitted SDK event is HMAC-SHA256 signed and chained to its predecessor. The optional atomic disk outbox persists it before enqueue and replays it after restart.

Deliver & Anchor

Ingest verifies and countersigns accepted events. Hosted records feed daily Merkle roots that can be checked against S3 Object Lock and Git anchors with the corresponding offline verifier.

THE DIFFERENCE

Without obsvr vs. with obsvr

Without obsvr

LLM calls happen with no visibility
PII flows freely into third-party APIs
No record of what actually happened
Policy enforcement is probabilistic at best
Audit prep takes weeks of manual log assembly
No way to prove compliance after the fact

With obsvr

Supported enforcing routes checked before provider or tool execution
Configured PII detected, redacted, or blocked in process
Denied tools tested at zero side-effect executions
Optional strict receipts committed before selected provider calls
Cryptographically signed trail for emitted events
Deterministic policy enforcement - no guessing
Tamper-evident continuity, plus optional device attribution
Signed deployment coverage names required bindings and explicit exclusions
Optional disk outbox replays signed records after restart

DATA HANDLING

What obsvr sees vs. what it doesn't

What reaches obsvr

Signed event metadata (timestamps, model, token count)
Policy evaluation results (PERMIT/DENY + reason)
Content hashes (SHA-256 of prompt and response)
Merkle tree position and chain anchors
Captured prompt/response values when storage policy permits

What stays local

API keys and credentials
Provider authentication
Caller-owned request objects
Content blocked or redacted before capture

Decision timeline

Governed LLM calls and agent runs, audited against an explicit boundary

audit_pipeline.log01/07

Intercept

Supported Node provider modules can be intercepted at process start; explicit wrappers govern the client reference your app uses

supported import hook · explicit wrappers elsewhere

What you get

Governance tooling for agent systems

Policy Engine & Management

Supported enforcing routes check rules before dispatch; response policy depends on the documented surface. Rules are versioned and content-addressed, and backend policy changes are attributed and sealed into the evidence stream.

Shadow mode
Org policy packs & inheritance
Signed policy-change history

Policy EvaluatorPROMPT RECEIVED

Keyword Block"SSN" in promptBLOCKED

PII Pre-callEmail detectedREDACTED

Topic Allow-listFinance ✓ALLOWED

Post-call Scan-SKIPPED

Evaluated in 0.3msBLOCKED

Agentic Governance

Obsvr extends call-level controls through supported model-start handlers, agent runners, governed tools, step limits, loop detection, delegation trackers, and final-output policy. Enforcing tool routes are tested at the side-effect boundary; tracing-only callbacks are labeled separately.

Tool restrictions
Step limits & runaway-loop detection
Kill switch

Agent Run Tracerun_cr_a1b2c3d4

CrewAI

3 / 10 stepsENFORCED

Investigate & Respond

When a policy fires or an agent misbehaves, obsvr turns the signed events into a structured incident - root cause, blast radius, and the exact next step - so you are not grepping logs at 2am. Pivot on any user, tool, or model, then export a signed incident report.