tps-0.1

Tervos Protocol

The organizational layer for agentic AI.

Identity. Policy. Audit. For autonomous AI organizations.

$ npx tervos init --template demo
✓ Created TERVOS.yaml
✓ Created .tervos/ directory

$ tervos start --dev
✓ Organization "Demo AI Corp" bootstrapped
✓ Agent ceo activated (did:key:z6Mk...)
✓ Agent analyst activated (did:key:z6Mk...)
✓ Agent writer activated (did:key:z6Mk...)
✓ 2 tasks queued

Get Started Read the Spec GitHub

AI agents can talk to tools (MCP) and to each other (A2A). They can't form organizations.

No standard identity. No policy enforcement. No audit trail. No compliance. Every team building multi-agent systems reinvents this from scratch.

🔑

Identity

Cryptographic DIDs for every agent. You know who did what, provably.

🛡️

Policy

Budget limits, approval gates, delegation rules. Enforced automatically, not suggested.

📋

Audit

Immutable event log of every decision. Full traceability. Export to your SIEM.

🏢

Organization

Roles, teams, reporting lines. Defined in YAML, enforced at runtime.

Architecture

Tervos sits on top of MCP and A2A, adding the organizational layer that neither protocol provides.

 Layer 8: Resilience v0.2+ 
 Layer 7: Observability 
 Layer 6: Knowledge 
 Layer 5: Economics 
 Layer 4: Organization v0.1 ✓ 
 Layer 3: Communication 
 Layer 2: Policy 
 Layer 1: Identity 
MCP
A2A

Your entire AI org in one file

TERVOS.yaml is the single source of truth for your autonomous organization.

# TERVOS.yaml — your entire AI org in one file
schema: tervos/v1
name: "My AI Company"

governance:
  autonomy_level: semi-autonomous
  budget:
    monthly_limit_usd: 5000

agents:
  ceo:
    role: "CEO"
    model: "claude-sonnet-4-20250514"
    capabilities: [strategy, delegation]
    budget_monthly_usd: 1000
    can_delegate_to: [engineer]

  engineer:
    role: "Engineer"
    reports_to: ceo
    model: "claude-sonnet-4-20250514"
    capabilities: [coding, testing]
    budget_monthly_usd: 2000

Already using Paperclip? Import your existing setup:

$ tervos import --from paperclip ./my-company.json
Imported 5 agents, 12 tasks. All agents set to supervised.

Built for the regulatory reality of 2026.

Ship AI agents without legal surprises.

NIST AI Agent Standards

Alignment built-in

EU AI Act

Progressive autonomy model

ISO 42001

Audit-ready event logs

Default-deny

Policy enforcement