AI Orchestration

Truthlocks AI Orchestration enables you to define, execute, and monitor multi-agent workflows with built-in cost attribution, safety guardrails, and complete audit trails.

Overview

┌─────────────────────────────────────────────────────────────┐
│                    Orchestration Engine                       │
│                                                               │
│  Workflow Definition → Step Execution → Result Aggregation   │
│         │                    │                    │           │
│         ▼                    ▼                    ▼           │
│   ┌──────────┐     ┌──────────────┐     ┌──────────────┐    │
│   │ Guardrail│     │  Agent Pool  │     │  Cost Tracker │    │
│   │  Engine  │     │              │     │              │    │
│   └──────────┘     │ Agent A      │     │ $0.12 total  │    │
│                    │ Agent B      │     │ 847 tokens   │    │
│                    │ Agent C      │     └──────────────┘    │
│                    └──────────────┘                          │
└─────────────────────────────────────────────────────────────┘

All orchestration and workflow endpoints have an interactive API playground. Open any endpoint in the Orchestrations & Workflows API reference and click Send to try it against the Sandbox.

Quick start

Execute an orchestration

curl -X POST https://api.truthlocks.com/v1/orchestrate/execute \
  -H "X-API-Key: $API_KEY" \
  -d '{
    "name": "quarterly-compliance-review",
    "agents": ["maip-agent:01JAAAA", "maip-agent:01JBBBB"],
    "steps": [
      {
        "name": "gather-evidence",
        "agent_id": "maip-agent:01JAAAA",
        "action": "compliance.gather",
        "input": { "scope": "soc2", "period": "Q1-2026" }
      },
      {
        "name": "evaluate-controls",
        "agent_id": "maip-agent:01JBBBB",
        "action": "compliance.evaluate",
        "depends_on": ["gather-evidence"],
        "input": { "framework": "soc2-type2" }
      }
    ],
    "guardrails": {
      "max_cost_usd": 5.00,
      "max_duration_seconds": 300,
      "require_trust_score_above": 70
    }
  }'

Response:

{
  "id": "maip-orch:01JOOOO",
  "status": "running",
  "steps": [
    { "name": "gather-evidence", "status": "running" },
    { "name": "evaluate-controls", "status": "pending" }
  ],
  "started_at": "2026-04-06T14:00:00Z"
}

Check Orchestration Status

curl https://api.truthlocks.com/v1/orchestrate/executions/maip-orch:01JOOOO \
  -H "X-API-Key: $API_KEY"

{
  "id": "maip-orch:01JOOOO",
  "status": "completed",
  "steps": [
    { "name": "gather-evidence", "status": "completed", "duration_ms": 12400 },
    { "name": "evaluate-controls", "status": "completed", "duration_ms": 8200 }
  ],
  "cost": {
    "total_usd": 1.24,
    "breakdown": [
      { "step": "gather-evidence", "usd": 0.87, "tokens": 12400 },
      { "step": "evaluate-controls", "usd": 0.37, "tokens": 5200 }
    ]
  },
  "completed_at": "2026-04-06T14:00:21Z"
}

LLM Inference Logging

Track every LLM call made during orchestration:

curl -X POST https://api.truthlocks.com/v1/orchestrate/llm/inference \
  -H "X-API-Key: $API_KEY" \
  -d '{
    "model": "claude-sonnet-4-6-20250514",
    "provider": "anthropic",
    "input_tokens": 1200,
    "output_tokens": 450,
    "latency_ms": 2400,
    "cost_usd": 0.012,
    "step_name": "evaluate-controls"
  }'

Workflows

For recurring orchestration patterns, define reusable workflows:

Create a Workflow

curl -X POST https://api.truthlocks.com/v1/orchestrate/workflows \
  -H "X-API-Key: $API_KEY" \
  -d '{
    "name": "weekly-risk-assessment",
    "description": "Automated weekly risk assessment across all vendors",
    "steps": [
      {
        "name": "scan-vendors",
        "action": "risk.scan",
        "input_template": { "vendor_list": "{{vendors}}" }
      },
      {
        "name": "score-risks",
        "action": "risk.score",
        "depends_on": ["scan-vendors"]
      },
      {
        "name": "generate-report",
        "action": "report.generate",
        "depends_on": ["score-risks"],
        "input_template": { "format": "pdf", "recipients": "{{team_email}}" }
      }
    ],
    "schedule": "0 9 * * MON",
    "guardrails": {
      "max_cost_usd": 10.00,
      "max_duration_seconds": 600
    }
  }'

Execute a Workflow

curl -X POST https://api.truthlocks.com/v1/orchestrate/workflows/maip-workflow:01JWWWW/execute \
  -H "X-API-Key: $API_KEY" \
  -d '{
    "variables": {
      "vendors": ["vendor_a", "vendor_b", "vendor_c"],
      "team_email": "risk-team@corp.com"
    }
  }'

Guardrails

Guardrail	Description
`max_cost_usd`	Hard cost ceiling — orchestration halts if exceeded
`max_duration_seconds`	Maximum wall-clock time
`max_tokens`	Total token budget across all LLM calls
`require_trust_score_above`	Minimum agent trust score to participate
`require_human_approval`	Steps that need human sign-off before proceeding
`allowed_models`	Whitelist of permitted LLM models
`blocked_actions`	Actions that are explicitly forbidden

Guardrail Check

curl -X POST https://api.truthlocks.com/v1/guardrails/check \
  -H "X-API-Key: $API_KEY" \
  -d '{
    "agent_id": "maip-agent:01JAAAA",
    "action": "orchestrations:execute",
    "context": {
      "estimated_cost_usd": 3.50,
      "models": ["claude-sonnet-4-6-20250514"],
      "duration_estimate_seconds": 120
    }
  }'

{
  "allowed": true,
  "checks": [
    { "rule": "max_cost_usd", "status": "pass", "limit": 5.00, "actual": 3.50 },
    { "rule": "allowed_models", "status": "pass" },
    { "rule": "trust_score", "status": "pass", "score": 87, "minimum": 70 }
  ]
}

Cost Attribution

All costs are attributed at multiple levels:

Per-step: Each step tracks its own token usage and cost
Per-agent: Aggregate cost across all steps an agent participated in
Per-orchestration: Total cost for the entire workflow execution
Per-tenant: Monthly rollup for billing purposes

Cost data feeds into the billing system and is visible in the console dashboard.

Billing & Quotas

Metric	Developer	Business	Enterprise
Orchestrations/mo	10	1,000	Unlimited
Workflow definitions	2	50	Unlimited
LLM inference logs/mo	100	10,000	Unlimited

Next steps

Machine identity

Full MAIP protocol overview and quick start.

Trust scores

Continuous behavioral trust evaluation for agents.

Event streaming

Generate receipts for messages flowing through Kafka, EventBridge, Kinesis, NATS, and Redis Streams.

AI model connectors

Automatic receipts for LangChain, LlamaIndex, CrewAI, and other AI framework calls.

Getting Started

Guides

Trust & Compliance

Security

Operations

Enterprise Packs

Machine Identity (MAIP)

Risk & Fraud

Specification

SDKs

AI Orchestration

AI Orchestration

Overview

Quick start

Execute an orchestration

Check Orchestration Status

LLM Inference Logging

Workflows

Create a Workflow

Execute a Workflow

Guardrails

Guardrail Check

Cost Attribution

Billing & Quotas

Next steps

Machine identity

Trust scores

Event streaming

AI model connectors

​AI Orchestration

​Overview

​Quick start

​Execute an orchestration

​Check Orchestration Status

​LLM Inference Logging

​Workflows

​Create a Workflow

​Execute a Workflow

​Guardrails

​Guardrail Check

​Cost Attribution

​Billing & Quotas

​Next steps

Machine identity

Trust scores

Event streaming

AI model connectors

AI Orchestration

Overview

Quick start

Execute an orchestration

Check Orchestration Status

LLM Inference Logging

Workflows

Create a Workflow

Execute a Workflow

Guardrails

Guardrail Check

Cost Attribution

Billing & Quotas

Next steps