Allow, block, redact, or require review before documented model and tool actions execute. Sign the same decision for later verification.
Apache-2.0 SDKs · local enforcement needs no hosted account
Pre-dispatch
blocks on documented enforcing routes
15+
documented integration surfaces
Zero executions
for denied tools on tested routes
Signed coverage
expected bindings and exclusions attested
Obsvr proof lab
Run three representative actions through a simplified local policy model. Watch the verdict, execution count, and evidence hash change together—without sending content anywhere.
Write your own payload in the full playgroundIntercepted call
{
"type": "tool.call.pre",
"tool": "sql_query",
"arguments": {
"query": "DROP TABLE customers;"
},
"agent": "data-agent"
}evaluating…
This page models selected public rules in your browser. The production SDK adds the full policy pipeline, authenticated HMAC chaining, optional device signatures, delivery accounting, and provider-specific enforcement boundaries.
The stakes
Three facts you can verify - a deadline, an attack path, and a leak. Obsvr addresses them through enforcement, investigation, and verifiable evidence.
Dec 2, 2027
EU AI Act - Annex III high-risk rules
Annex III rules apply from Dec 2027; high-risk AI embedded in regulated products follows in Aug 2028. Art. 12 logging cannot be backfilled after deployment.
source: Regulation (EU) 2026/174490+
organizations hit by prompt injection
In 2025 alone. It is now the #1 risk on OWASP's Top 10 for LLM applications - and it arrives as ordinary input your agent is trusted to read.
source: CrowdStrike 2026 Global Threat Report34.8%
of data sent to AI is now sensitive
The cited workplace dataset found over a third of enterprise data shared with AI was sensitive. Your own exposure depends on your traffic and controls.
source: Cyberhaven 2025 AI Adoption & Risk ReportFor applied AI teams
Once AI can trigger real work, the team shipping it needs a policy decision before execution and a verifiable decision record after it.
A log can describe the email, refund, export, or tool call after the side effect is real.
Detecting sensitive content after a provider or tool receives it is not prevention.
Ordinary logs rarely preserve the exact actor, model, policy version, reason, and event order together.
From 2 December 2027, Annex III high-risk systems enter the deferred rules window, including lifecycle logging obligations. Turn that on late and the gap is permanent.
See the article-by-article coverageThe gaps
Not an observability tool. Not a trace viewer.
Logs record the violation after it happened. They can't block the call.
Output filters clean up after generation. They never stop what was sent.
Dashboards show what happened. They can't produce a record an auditor can verify.
A detector only sees the traffic that reached it. It can't tell you about the call that bypassed it entirely.
obsvr = control + enforcement + proof + coverage attestation
What obsvr does
Define versioned rules, approvals, and workload boundaries once. Replay a candidate against recorded inputs, review what changes, then promote the exact content-addressed policy you approved.
Explore SecureSee it in action
Edit policies from the dashboard and live SDKs pick them up in minutes without a redeploy. Governed events stay linked to the rules and context used for investigation.
Events captured
0
Open incidents
0
Controls passing
3 / 3
Live event stream
Incident board
Measured, not marketed
24.7–25.1µs p50 across two complete 0.11.2 benchmark passes.
63.2–64.1µs p50 with rules, PII scan, quotas, hooks, signing, and enqueue.
524.3–525.1µs p50 under the same local-policy benchmark shape.
Two complete passes verified retained signatures, chain links, accounting, and declared loss.
Reproducible: BENCHMARKS.md · raw evidence retained for both runs
Integrations
Supported provider, framework, and protocol routes produce one decision format. The coverage grade tells you exactly where policy runs before dispatch and where a separate gate is needed.
Where obsvr fits
obsvr is not a replacement for enterprise governance, gateways, or security platforms. It adds enforceable product policy and decision evidence on its documented integration paths.
FAQ
8 questions
12
Contact
Bring one provider, tool, agent, or MCP route. We will map its enforce-or-observe boundary, run the public SDK locally, and show which optional hosted evidence workflows are available in beta.