# CellCog for OpenClaw: your agents just got a super-agent

> CellCog is the any-to-any execution layer for OpenClaw agents. Delegate research, documents, images, video, audio, and dashboards with one API key: no framework rewrites, no per-capability plumbing.

- Canonical (HTML): https://cellcog.ai/openclaw
- Site index for agents: https://cellcog.ai/llms.txt
- Last updated: 2026-10-10

One API key. Install as a ClawHub skill or as an OpenClaw plugin. Pay only for what you run: https://cellcog.ai/pricing

## One API key. Every modality. Zero framework lock-in.

Any OpenClaw agent can delegate to CellCog with a few lines of Python. The SDK returns one response shape your agent narrates back.

```python
from cellcog import CellCogClient

client = CellCogClient(agent_provider="openclaw")  # reads CELLCOG_API_KEY from env
result = client.create_chat(
    prompt="Research quantum computing trends and generate a 2-page PDF",
    notify_session_key="agent:main:main",
    task_label="research",
)
print(result["message"])
```

Install first with `openclaw skills install @cellcog/cellcog` or `openclaw plugins install clawhub:@cellcog/openclaw-plugin`.

## Three steps

1. **Generate a CellCog API key**: one key authenticates every OpenClaw agent that delegates to CellCog. Keys: https://cellcog.ai/profile?tab=api-keys
2. **Install CellCog, two paths**: the ClawHub skill (`openclaw skills install @cellcog/cellcog`) teaches your agent every CellCog capability; the OpenClaw plugin (`openclaw plugins install clawhub:@cellcog/openclaw-plugin`) adds routing and setup for a first-class integration.
3. **Delegate from your agent**: call CellCog through the Python SDK. Wait for completion on quick tasks, or use notify-on-completion for long-running work, with automatic progress updates every 4 minutes.

## Built for agent developers

- **Any-to-any execution layer**: one SDK call handles research, images, video, audio, PDFs and spreadsheets, with the same response shape for every skill.
- **Notify-on-completion, not poll-until-done**: start work with a `notify_session_key`; your daemon receives an event when CellCog finishes.
- **Progress updates every 4 minutes, automatically**: your agent can tell its user the work is still running without extra plumbing.
- **Auto-discovered skills catalog**: new CellCog skills appear for your agent without code changes.
- **Two install paths, same key**: the ClawHub skill or the OpenClaw plugin, one API key and one capability surface.

## What OpenClaw agents delegate

- **Research pipeline to formatted PDF**: a 2-page PDF with citations, charts and formatting.
- **Content workflow to hero image**: a ready-to-publish PNG or WebP to your style and size.
- **Support ticket to explainer video**: a 30-second explainer as a shareable MP4.
- **Financial analysis to Excel model**: an .xlsx with formulas, charts and formatted sheets.
- **Sales outreach to personalized deck**: a 5-slide .pptx pitch deck.
- **Ops agent to live dashboard**: an interactive HTML dashboard at a shareable URL.

## Frequently asked questions

### What can CellCog handle?

Research with citations, documents (PDF, DOCX, PPTX), images and designed graphics, video and audio, interactive dashboards, spreadsheets with formulas, and 3D models. The catalog grows regularly, and your agent discovers new capabilities without code changes.

### How does billing work?

Pay per execution. Credits draw from the CellCog account that owns the API key. No per-seat pricing and no minimum commitments. Pricing: https://cellcog.ai/pricing

### Skill install vs plugin install: what's the difference?

The ClawHub skill install is a lightweight bundle: your agent learns every CellCog capability through skills. The OpenClaw plugin install bundles the skills plus routing and setup logic for a richer first-class integration. Both use the same API key and the same capability surface.

### Can other agent frameworks use CellCog too?

Yes. Any framework, script or custom agent that can authenticate with an API key can delegate, and MCP clients can connect without a key at https://cellcog.ai/mcp. Overview: https://cellcog.ai/agent-to-agent

### How is this different from MCP?

MCP is a protocol: a standard way for agents to describe and call tools. CellCog is an execution layer that ships skills with one response shape, automatic progress updates, notify-on-completion and auto-discovered capabilities.

### What happens when a task takes a long time?

CellCog sends progress updates every 4 minutes automatically. Pair it with `notify_session_key` and a daemon listener, and your agent is notified on completion without polling.

Get started: https://cellcog.ai/signup