Give AI agents, workloads, and devices a verifiable identity. Score their trust, make real-time trust decisions, and revoke compromised identities from one platform.
AI agents and compute infrastructure increasingly act autonomously: accessing resources, executing workloads, and making decisions without a human in the loop for every step.
01
Who is it?
02
What is it?
03
Can I trust it?
04
What is it allowed to do?
05
What evidence supports that decision?
06
What happens if trust changes?
ComputeID answers these questions continuously, not once at onboarding, and not by trusting a bearer token to keep telling the truth.
One trust layer, multiple primitives
ComputeID as a single system, not three disconnected products.
Identity
Passports
AgentPassport · DevicePassport · WorkloadPassport
Cryptographically identify agents, devices, and compute workloads.
Trust
Trust Score
Identity · Authority · Integrity · Evidence
Continuously evaluate trust using identity, ownership, evidence and behavior signals.
Turn trust signals into an actionable decision on every call, in real time.
Control
Revocation & Escrow
Kill switch · policy-driven escrow
Control access, assets and execution, and revoke when trust changes. Open streaming sessions are re-checked about every 7 seconds.
How a trust decision is made
Four signals in. One decision out.
Pick a scenario and watch the same request resolve differently. The demo is illustrative: it represents the authorization decision and permitted status that the Trust API returns for an action.
POST /v1/trust-decision · illustrative decision trace
Inbound signals
Identity98
ML-DSA-65 signature verified against issued passport
Authority95
write:refunds is in the agent’s declared scope
Integrity92
Call pattern consistent with 14-day baseline
Evidence90
Device attestation present and current
Illustrative Trust Decision Response
passport_id: agent-47 · action: write:refunds
decision: ALLOW
permitted: true
Trust Score
94
/ 100
Revoke
Deny
Quarantine
Restrict
Monitor
Allow
Decision
ALLOW
Passport valid and signature verified. The requested action falls inside the capabilities the agent was issued.
illustrative example
Illustrative traces. Scores, signals and field values are examples, not live customer data.
Architecture
The ComputeID Trust Stack.
Identity tells you who something is. Trust tells you what it should be allowed to do. ComputeID connects the two.
Decision layer
Trust Decisions
AllowAllow + monitorRestrictQuarantineDenyRevoke
Scoring layer
Trust Score
Identity
Agent Passport
Identity
Device Passport
Identity
Workload Passport
Evidence & Proof
Escrow
AI Systems + Compute Infrastructure
ComputeID doesn't just identify AI systems. It helps determine what they can be trusted to do. AgentPassport is live in production. The device-identity tier, WorkloadPassport and escrow are the newer layers of this same stack.
Built. Tested. Verifiable.
What's built today.
Built and covered by tests. Each line says how far it extends today.
Trust Score
Built and covered by automated tests.
Built
Reason + Evidence
Every Trust Score response returns the reasoning and evidence behind the decision, not just a number.
GPU clouds · neoclouds · data centers · compute marketplaces
Know which device or workload is actually calling before you bill it or hand it a credential.
Agentic AI
Agent platforms · autonomous agents · multi-agent systems
Give every agent a verifiable identity and a live, revocable trust decision on every action.
Enterprise
CISOs · AI platform teams · security & risk teams
Replace "we trust the API key" with a signed identity, a scored decision, and an audit trail.
Regulated & Sovereign
Government · defense · financial services · sovereign AI
The CNSA 2.0-tier passport (ML-DSA-87) extends the same primitive to devices.
Standards, not marketing
We name the exact primitives, not invented crypto brand names.
FIPS 204 · ML-DSA-65 signing
CNSA 2.0 · ML-DSA-87 tier
Hybrid · RSA-2048 + post-quantum
Get started
Issue your first passport in a few lines.
# pip install computeid-sdk
from computeid import ComputeIDClient, AgentPassport
client = ComputeIDClient(api_key="YOUR_API_KEY")
passport = AgentPassport.issue(
client, "ResearchAgent",
organization="Acme Corp",
capabilities=["read", "web_browse"],
)
# a live verify call, every time
if passport.is_trusted():
run_your_agent(passport)
passport.revoke(reason="Unexpected behaviour detected")
01
Issue a passport
Register the agent with the capabilities it is allowed to use.
02
Scope its capabilities
A passport is issued for exactly what the agent is allowed to do, nothing broader.
03
Check trust before it acts
Verify the passport live, or ask the Trust API for a score and decision for a specific action.
04
Revoke on suspicion
One call, and later verifications of that passport fail.
The ComputeID Platform
Five layers. One trust decision.
Passports
Identity for agents, devices and workloads.
Trust Score
Quantifies trust using evidence and live signals, not a static rating.
Trust Decisions
Converts trust into an actionable policy outcome: allow, allow and monitor, restrict, quarantine, deny or revoke.
Escrow
Controls access, execution and assets with policy-driven conditions.
Control
Restrict, revoke and respond when trust changes. Passport checks are live on every call, and open streaming sessions are re-checked about every 7 seconds.
Identity tells you who something is. Trust tells you what it should be allowed to do.
ComputeID connects the two.
Honest FAQ
We'll tell you what's still narrow, not just what's built.
No. We do not hold a SOC 2 report today. Every passport verification returns a CA-signed receipt that you can verify independently. We'll say so plainly if that changes.
Pilot program
A focused engagement against your real environment.
Pilot program
Custom pricingContact sales
A focused deployment to evaluate ComputeID against your real AI or compute environment: scoped and priced with you.
Environment assessment
Identity integration
Agent / workload / device onboarding
Trust Score configuration
Trust Decision policies
Escrow configuration, where applicable
Integration support
Evidence and audit review
Final pilot report + production deployment recommendations