Before your agent moves the money,it asks DecisionLens.
Real-time regulatory verdicts for AI agents — sub-100ms, with the full reasoning chain your auditor will ask for.
- MCP server & REST
- Works with Claude Code, Cursor, LangChain, OpenAI Agents SDK
- Policy templates: AML, KYC, access control, fair lending
- SOC 2 in progress
curl -X POST https://api.llmforce.ai/v1/evaluate \
-H "Authorization: Bearer dl_live_xxx" \
-H "Content-Type: application/json" \
-d '{
"policy_context": "aml_standard_v4",
"decision_type": "wire_transfer",
"case": {
"amount_usd": 9850,
"originator_id": "cust_8f21",
"beneficiary_country": "MT",
"prior_24h_count": 3,
"prior_24h_total_usd": 28400
},
"proposed_action": "APPROVE"
}'Four steps. One API call. Auditor-ready.
Agent decides
Your agent is about to act — approve a wire, grant access, onboard a customer, extend credit.
Agent calls /evaluate
One POST with the case and a policy slug. MCP server or REST. 60-second install.
DecisionLens evaluates
Policy match, risk scoring, full reasoning chain — returned in under 100ms.
Agent proceeds, denies, or escalates
Verdict in hand. Audit trace stored. Your regulator gets the receipts, not a black box.
Plug in a policy. Or bring your own.
Start with a template — extend it with your firm’s thresholds, jurisdictions, and override rules. Every template is versioned, diffable, and citable from a verdict.
Read the policy specThree reasons it’s worth one extra API call.
Real-time, not retrospective.
Compliance happens before the action — not in next quarter’s audit log. p95 under 100ms.
Reasoning chains, not approvals.
Every verdict ships with a step-by-step trace, policy citations, and confidence. Auditor-ready by default.
Agent-native.
MCP server, REST, and a Claude Code skill. Drop it into any agent loop without re-plumbing.
Every verdict cites the policy, the clause, and the case data that triggered it. No black-box approvals.
Ship agents your compliance officer trusts.
Free tier: 1,000 evaluations / month. No credit card.