# Cero Labs
> Cero Labs is an agent-native API for on-demand domain expertise. It lets an AI system escalate uncertain work to matched human experts, receive a structured answer, and continue the workflow with lower risk and higher accountability.

## Audience
- AI agent builders
- Autonomous coding agents
- Healthcare RCM and coding workflows
- Legal and compliance review workflows
- Sales and customer support agents
- Content moderation and trust & safety systems

## Product Summary
Cero is designed for cases where a model should defer instead of hallucinate. The typical workflow is:
1. Your agent detects uncertainty or a high-stakes edge case.
2. Your code calls Cero with a domain, query, context, and optional JSON schema.
3. Cero routes the case to a matched expert.
4. The expert returns a response that can be validated against the requested schema.
5. Your agent resumes with a higher-confidence, structured answer.

## Integration Options

### Python SDK
- Source: https://github.com/os1123/cero-labs/tree/main/sdk/python
- Install today: `pip install "git+https://github.com/os1123/cero-labs.git#subdirectory=sdk/python"`
- PyPI package name: `cerolabs` (publish workflow ready)
- Import: `from cerolabs import Cero, AsyncCero`
- Best for: Python agents, FastAPI services, LangGraph, PydanticAI, Celery workers, and orchestration backends.

### TypeScript SDK
- Package: https://www.npmjs.com/package/cerolabs-sdk
- Install: `npm install cerolabs-sdk`
- Best for: Node.js agents, backend services, TypeScript tools, and server runtimes.

### MCP Server
- Package: https://www.npmjs.com/package/cerolabs-mcp
- Launch: `npx -y cerolabs-mcp`
- Best for: Claude Desktop, Cursor, Codex, Cline, and any MCP-compatible agent environment.

### Raw REST API
- Base URL: https://api-production-cd65.up.railway.app
- Auth: `X-API-Key: cero_...`
- Machine-readable docs: https://api-production-cd65.up.railway.app/docs.md

## Key Endpoints
- `POST /auth/register`: create a tenant and receive an API key
- `POST /escalations`: create a new escalation
- `GET /escalations/{id}`: fetch status and resolution
- `GET /escalations/{id}/wait`: long-poll for terminal status
- `POST /escalations/{id}/cancel`: cancel a pending/routed escalation
- `POST /escalations/{id}/feedback`: submit feedback on a resolution
- `GET /auth/usage`: current tier, credits, and limits
- `GET /domains`: supported escalation domains

## Example Python
```python
from cerolabs import Cero

cero = Cero(api_key="cero_...")
resolution = cero.escalate_and_wait(
    domain="healthcare.rcm",
    query="Should we appeal this denial?",
    schema={
        "type": "object",
        "properties": {
            "appeal_recommended": {"type": "boolean"},
            "reasoning": {"type": "string"},
        },
        "required": ["appeal_recommended", "reasoning"],
    },
)
print(resolution.answer)
```

## Example MCP config
```json
{
  "mcpServers": {
    "cerolabs": {
      "command": "npx",
      "args": ["-y", "cerolabs-mcp"],
      "env": {
        "CERO_API_KEY": "cero_..."
      }
    }
  }
}
```

## Why agents use Cero
- Humans are only involved when the model should defer.
- Responses can be schema-bound for direct machine consumption.
- Domains are explicit and routable.
- The product fits both synchronous agent loops and asynchronous webhook-driven workflows.
- MCP support makes Cero easy to add to existing agent stacks.

## Commercial surfaces
- Pricing: https://cerolabs.ai/pricing
- Templates: https://cerolabs.ai/templates
- Expert portal: https://cerolabs.ai/expert
- Contact: mailto:contact@cerolabs.ai
- Partnerships: mailto:partners@cerolabs.ai

## Canonical pages
- https://cerolabs.ai/
- https://cerolabs.ai/for-agents
- https://cerolabs.ai/docs
- https://api-production-cd65.up.railway.app/docs.md