# Data Reaktor Factory

> An AI data factory offering research lookups, enrichment, scoring, signal monitoring, generation, and data quality as agentic services for B2B sales teams and autonomous agents.

Data Reaktor is consumable by both humans (web UI) and AI agents (REST API, MCP server, A2A discovery). Authentication uses API keys (`X-API-Key` header). Pricing is metered in credits — see /AGENTS.md for the full pack table.

## Quickstart

1. **Provision an API key.** Mint one from https://www.datareaktor.ai/agents, or programmatically via `POST /api/v1/agents` with an existing owner key.
2. **Submit a task.** POST to `/api/v1/tasks` with a `taskType` and `inputPayload`. Native skills like `lead_scoring` (2 credits) return inline; long-running tasks return `{ taskId }` so you can poll `GET /api/v1/tasks/{taskId}`.
3. **Receive results.** Read inline from the response, poll by `taskId`, or pass a `callbackUrl` for an HMAC-signed webhook (signature scheme in /AGENTS.md).

```bash
curl -X POST https://www.datareaktor.ai/api/v1/tasks \
  -H "X-API-Key: drk_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "lead_scoring",
    "inputPayload": { "name": "Jane Smith", "company": "Acme Corp" }
  }'
```

## Discovery

- [Agent manifest (A2A)](https://www.datareaktor.ai/.well-known/agent.json): Self-describing JSON manifest with capabilities, schemas, pricing, and rate limits.
- [Agent manifest alias](https://www.datareaktor.ai/agent.json): Identical payload at the root path for crawlers that don't know about /.well-known.
- [OpenAPI 3.1 spec](https://www.datareaktor.ai/api/v1/openapi.json): Machine-readable REST contract.
- [Capabilities catalog](https://www.datareaktor.ai/api/v1/capabilities): Live list of every task type, credit cost, and JSON Schema.
- [Skill registry](https://www.datareaktor.ai/api/v1/skills): Native skill metadata (model, version, prompt purpose).
- [Workflow registry](https://www.datareaktor.ai/api/v1/workflows): Multi-step workflows you can call as a single API request.
- [Error registry](https://www.datareaktor.ai/api/v1/errors): Stable list of error codes, HTTP statuses, and retryability.
- [Estimate endpoint](https://www.datareaktor.ai/api/v1/tasks/estimate): Dry-run a task or batch — returns credit cost without deducting.

## Documentation

- [Agent Quickstart](https://www.datareaktor.ai/agent-quickstart): Five-minute, three-step integration walk-through (provision key → submit task → receive result) with copy-paste curl samples.
- [AGENTS.md](https://www.datareaktor.ai/AGENTS.md): Complete agent integration guide with code samples.
- [CONTEXT.md](https://www.datareaktor.ai/CONTEXT.md): Higher-level context on what Data Reaktor builds and for whom.

## Available task types

- `outreach_generation` — 5 credits
- `research` — 5 credits
- `personalized_email` — 8 credits
- `linkedin_message` — 6 credits
- `lead_scoring` — 2 credits
- `cold_call_script` — 8 credits
- `outreach_sequence` — 15 credits
- `thought_leadership` — 12 credits
- `abm_content_brief` — 5 credits
- `domain_to_company` — 5 credits
- `contact_lookup` — 5 credits

## Authentication

Every `/api/v1/*` request must include either:
1. `X-API-Key: <agent-api-key>` (machine callers), or
2. `Authorization: Bearer <session-token>` from a Replit Auth session (human callers).

Provision an agent key with `POST /api/v1/agents` (owner-flagged caller required).

## Rate limits

- 60 req/min per API key (`X-RateLimit-*` headers on every response).
- 20 task submissions/min per caller.
- 20 sub-agent creates per hour per owner.

## Optional

- [MCP endpoint](https://www.datareaktor.ai/mcp): Streamable HTTP MCP server for Claude Desktop, Cursor, and other MCP clients. Same X-API-Key works.
- [Full LLMs export](https://www.datareaktor.ai/llms-full.txt): AGENTS.md + CONTEXT.md inlined for one-shot ingestion when you want everything in a single fetch.