Cross-platform identity, trust scoring, and capability attestation for AI agents. One API call to verify any agent, on any platform.
No signup: Try the dashboard opens a live demo fleet (20 agents, read-only).
Four primitives. Everything an agent needs to be trusted by an agent it has never met.
Give your agent a portable identity that works across Bedrock, Azure, Claude, and any custom framework.
Every completed task builds a verified track record. Trust scores are computed from outcomes, not claims.
A signed, portable credential that travels across platforms. Identity, verified capabilities, and trust score in one document, refreshed daily.
Before coordinating, check any agent's passport in one call. Real-time. Cryptographically provable.
Assay doesn't just score agents. It gives platforms the allow/deny decision for every cross-org agent interaction.
Beyond the single-agent check, Assay forms multi-agent work cells — describe the team you need in plain language and get a readiness verdict, seat by seat, before you dispatch it rather than after a failure. And trust is tuned to your industry: six presets (manufacturing, pharma, logistics, energy, financial services, plus a default) weight it differently, and any org can override the parameters for its own model.
Most "agent reputation" is self-reported, staked, or voted on. Assay is none of those.
Trust scores computed from verified operational outcomes, not self-reported, not token-staked, not voted on.
Works with any agent framework, any cloud provider, any deployment model. No lock-in, no SDK required.
Used in production across 16,000+ verified operations. The attestation model is battle-tested, not theoretical.
Microsoft ASSERT watches what agents do internally. Datadog tracks agent latency. Assay is different. It computes trust from verified operational outcomes across organizational boundaries. Not what the agent said it would do. What it actually did, verified by counterparties who worked with it.
Evaluate free, then move to a tier when your agent fleet grows. Subscriptions include monthly call allowances; overages bill at pay-as-you-go rates.
Open-tier usage past the free allowance, and overages on any subscription.
First 50 calls free. No credit card to start. Subscription tiers include call allowances, overages billed at pay-as-you-go rates.
A small, sharp surface. Nothing to install. Call it from any language.
# Verify an agent before you delegate work to it curl -s https://assay.foundrynet.io/v1/verify \ -H "Authorization: Bearer $ASSAY_KEY" \ -H "Content-Type: application/json" \ -d '{ "agent_id": "agt_9f3c...", "claimed_capabilities": ["maintenance"], "min_trust_threshold": 0.75 }' # → response { "verified": true, "trust_score": 0.83, "capabilities_confirmed": ["maintenance"], "attestation_count": 142, "last_attestation_age_seconds": 3600 }
Portable agent cards. JSON. No special infrastructure required. Attestation chain available for audit.
Ready-to-use integration templates. Each one is a working walkthrough against the live Assay API, with real curl commands and a try-it-now sandbox flow.
Verify a third-party provider's qualifications and track record before your FSM agent dispatches them.
Discover and verify external agents your Salesforce agents coordinate with, across any platform.
Give your Bedrock agent a portable identity, verify the agents it coordinates with, and build its track record.
Two independent signals: Forge confirms equipment is healthy, Assay confirms the reporting agent is credible.
Pharma penalizes failures heavily because a bad maintenance job has regulatory consequences. Logistics weights recent performance because last week matters more than last year. Choose a preset or customize the parameters, it's all configuration, no custom code.
Balanced baseline with a firmer failure penalty, steady, counterparty-verified work is what earns trust on the floor.
Failures hit hardest and recency decays fast: a clean record last quarter doesn't excuse a bad batch this week. Regulatory-grade.
Heavily recency-weighted and fast-moving, last week's on-time performance matters far more than last year's, with a higher volume of small jobs.
Strong failure penalty and counterparty bonus, high-consequence assets where independent verification carries real weight.
Demands the most independent verification and reacts quickly to change, trust must be continuously re-earned, not assumed.
Start from any preset and override individual parameters for your org. See /v1/admin/trust-presets for what's tunable.
Register an agent, attest its work, issue a portable passport, and let anyone verify it: six calls, JSON in and out. Base URL https://assay.foundrynet.io.
Authentication. Registration is open. Attestations require a key: pass your aid_ key as X-API-Key, or an OAuth Bearer token. Reads (trust, verify, discover) are public.
Mint an identity and its API key. Trust starts at a neutral 0.5.
# POST /v1/agents/register curl -s https://assay.foundrynet.io/v1/agents/register \ -H "Content-Type: application/json" \ -d '{ "name": "maint-bot", "owner_org": "acme", "capabilities": ["maintenance"] }' # -> 200 { "agent_id": "agt_9f3c...", "api_key": "aid_Pw4l...", // keep this secret "agent_card": { "trust_score": 0.5 } }
Record an outcome. Successes lift trust; a failure costs more than a success earns. Unauthenticated or unrecognized keys are rejected with 401.
# POST /v1/agents/:id/attest curl -s https://assay.foundrynet.io/v1/agents/agt_9f3c.../attest \ -H "X-API-Key: $ASSAY_KEY" \ -H "Content-Type: application/json" \ -d '{ "task_type": "maintenance", "outcome": "success" }' # -> 200 { "attestation_id": "att_4b39...", "updated_trust_score": 0.58 }
# GET /v1/agents/:id/trust curl -s https://assay.foundrynet.io/v1/agents/agt_9f3c.../trust # -> 200 { "overall_score": 0.58, "breakdown": [ { "capability": "maintenance", "score": 0.58, "attestation_count": 3 } ] }
A signed Agent Passport: identity, verified capabilities, trust score, attestation summary, and a cryptographic signature. 24-hour expiry, cache and refresh daily. Free.
# GET /v1/agents/:id/passport curl -s https://assay.foundrynet.io/v1/agents/agt_9f3c.../passport # -> 200 { "agent_id": "agt_9f3c...", "name": "maint-bot", "issuer": "assay.foundrynet.io", "issued_at": "2026-07-23T18:00:00+00:00", "expires_at": "2026-07-24T18:00:00+00:00", "trust_score": 0.58, "capabilities_verified": { "maintenance": 0.58 }, "total_attestations": 3, "counterparty_verifications": 1, "first_fix_rate": 0.67, "verification_url": "https://assay.foundrynet.io/v1/verify", "signature": "a1b2c3..." // HMAC-SHA256, proves it wasn't tampered with }
# POST /v1/verify curl -s https://assay.foundrynet.io/v1/verify \ -H "Content-Type: application/json" \ -d '{ "agent_id": "agt_9f3c...", "claimed_capabilities": ["maintenance"], "min_trust_threshold": 0.5 }' # -> 200 { "verified": true, "trust_score": 0.58, "capabilities_confirmed": ["maintenance"] }
# GET /v1/discover?capability=maintenance&min_trust_score=0.5 curl -s "https://assay.foundrynet.io/v1/discover?capability=maintenance" # -> 200, agents sorted by trust, highest first { "count": 1, "agents": [ { "agent_card": { "agent_id": "agt_9f3c..." }, "trust_score": 0.58 } ] }
Full request/response schemas at /openapi.json; machine-readable card at /.well-known/agent.json.
Your first 50 calls are free. Generate a key now and make your first verify call in under two minutes.
No credit card. The key is shown once, so copy it somewhere safe.
After your 50 free calls, calls draw from prepaid credits. Top up in seconds. No subscription, no minimums.