Documentation

Basic API and integration documentation.

Enough to evaluate the contract, run a sample request and understand how authentication, tenancy, idempotency and webhooks behave. Full reference documentation is issued per engagement.

Integration guide Quick start

Full reference

Integration guide

The complete client guide: the three integration patterns, the onboarding path, authentication and scopes, idempotency and error handling, every live agent’s endpoints and payloads, webhook signature verification, building a connector for your own systems, security and data handling, and a go-live checklist.

Read the guide

Quick start

Nine steps to a first response.

1Choose an agent.
2Create or identify a tenant.
3Configure authentication.
4Review the agent API contract.
5Connect a mock or sandbox system.
6Submit a sample request.
7Review the response, tool actions and audit event.
8Configure approval rules.
9Move to pilot integration.

API conventions

Rules that hold across every agent.

CrewTAC AI Products API conventions
Convention	Standard
Base path	/v1
Content type	application/json unless file upload is required
Authentication	X-API-Key header, or OAuth where supported
Tenant context	Derived from the key; never sent by the client
Traceability	X-Correlation-ID on every request and response
Idempotency	Idempotency-Key header on every write
Errors	code + message + retryable + correlation_id
Webhooks	Signed, versioned, retry-safe

Credentials

API keys and OAuth client secrets are issued per tenant and per environment. They are used server-side only — never embed them in a browser, mobile bundle or public repository.

Worked example

Sales agent request and response.

The same shape applies across agents: a typed request in, a scored and reasoned result out, with an audit event written behind it.

POST /v1/leadsrequest

{
  "name": "Alex Morgan",
  "email": "alex@example.com",
  "company": "Example SaaS",
  "role": "COO",
  "company_size": "51-200",
  "industry": "SaaS",
  "source": "website",
  "consent_status": "granted",
  "problem": "We want to automate customer onboarding."
}

201 Createdresponse

{
  "lead": { "id": "lead_a41f...", "status": "qualified" },
  "qualification": {
    "score": 74,
    "fit": "high",
    "intent": "buying",
    "confidence": 0.71,
    "reason_codes": [
      { "code": "COMPANY_SIZE", "label": "Company size 51-200.", "points": 12 },
      { "code": "SENIOR_ROLE",  "label": "Decision-maker role.", "points": 9 },
      { "code": "INTENT",       "label": "Intent: buying.",      "points": 18 }
    ],
    "rules_version": "v1",
    "prompt_version": "sales.lead_classify@v1"
  },
  "actions": [
    { "action_type": "book_discovery_call", "status": "pending_approval" }
  ]
}

Agent endpoints

The /v1 surface, per agent.

AI Sales Qualification Agent

Agent overview

POST/v1/leadsSubmit a lead for qualification and scoring.
GET/v1/leads/{id}Retrieve a lead with its score, reason codes and status.
POST/v1/leads/{id}/approveApprove a recommended action at a human gate.
POST/v1/webhooks/crmReceive CRM-side changes to keep records in sync.

AI Customer Support Agent

Agent overview

POST/v1/support/messagesSend a customer message and receive a grounded answer or escalation.
POST/v1/knowledge/documentsIngest or update a document in the tenant knowledge base.
POST/v1/ticketsCreate or update a ticket in the connected ticketing system.
POST/v1/feedbackRecord thumbs-up/down and free-text feedback on an answer.

AI Document Processing Agent

Agent overview

POST/v1/documentsUpload a document for classification and extraction.
GET/v1/documents/{id}Retrieve extraction results with per-field confidence.
GET/v1/documents/schemasList the document types and fields this deployment extracts.
GET/v1/documents/review-queueOpen review tasks, with the reason and the fields to check.
POST/v1/documents/{id}/reviewSubmit human review corrections for low-confidence fields.
POST/v1/documents/{id}/exportDeliver validated structured data downstream.

AI Operations Workflow Agent

Agent overview

POST/v1/tasksSubmit an operational request for workflow execution.
GET/v1/tasks/{id}Inspect workflow state, step history and pending approvals.
POST/v1/tasks/{id}/approveClear a human approval gate for a held step.
GET/v1/workflowsList the versioned workflow definitions this deployment runs.
GET/v1/toolsList every action the agent can take, with schema and scope.
POST/v1/webhooks/eventsReceive an external event that should start a workflow.

AI CRM Intelligence Agent

Agent overview

POST/v1/crm/askAsk a natural-language question scoped to the caller permissions.
POST/v1/crm/actions/previewPreview the exact CRM writes an action would perform.
POST/v1/crm/actions/executeExecute a previewed action after approval.
GET/v1/crm/accounts/{id}Retrieve an account summary with supporting evidence.

AI Legacy Modernization Agent

Agent overview

POST/v1/assessmentsOpen an assessment for a legacy application or portfolio.
POST/v1/assessments/{id}/artifactsAttach sanitized artifacts for secure analysis.
GET/v1/assessments/{id}/findingsRetrieve findings with the evidence behind each one.
POST/v1/assessments/{id}/roadmapGenerate a prioritized, phased modernization roadmap.

Evaluation

What we measure before and after changes.

Prompt and model changes go through the same regression gate as code.

Golden test cases
Structured-output accuracy
Unsupported-answer / hallucination rate where applicable
Tool-call success rate
Latency
Token / model cost
Authorization tests
Approval-gate tests
Regression tests after model or prompt changes

Next step

Need the full API reference?

Sandbox credentials, environment details and complete reference documentation are issued as part of a pilot engagement.

Request API access See the integration process

Or email sales@crewtac.com