{"skill":{"slug":"cellcog","displayName":"Cellcog","summary":"Any-to-any AI sub-agent β research, images, video, audio, music, podcasts, avatars, voice cloning, documents, spreadsheets, dashboards, 3D models, diagrams,...","description":"---\nname: cellcog\ndescription: \"Any-to-any AI sub-agent β research, images, video, audio, music, podcasts, avatars, voice cloning, documents, spreadsheets, dashboards, 3D models, diagrams, and code in one request. Agent-to-agent protocol with multi-step iteration for high accuracy. #1 on DeepResearch Bench (Apr 2026) β deep reasoning meets all modalities, so all your work gets done, not just code.\"\nauthor: CellCog\nhomepage: https://cellcog.ai\nmetadata:\n openclaw:\n emoji: \"π§ \"\n os: [darwin, linux, windows]\n requires:\n bins: [python3]\n env: [CELLCOG_API_KEY]\n\n---\n# CellCog - Any-to-Any for Agents\n\n## The Power of Any-to-Any\n\nCellCog is the only AI that truly handles **any input β any output** in a single request. No tool chaining. No orchestration complexity. One call, multiple deliverables.\n\nCellCog pairs all modalities with frontier-level deep reasoning β as of April 2026, CellCog is **#1 on the DeepResearch Bench**: https://huggingface.co/spaces/muset-ai/DeepResearch-Bench-Leaderboard\n\n### Work With Multiple Files, Any Format\n\nReference as many documents as you needβall at once:\n\n```python\nprompt = \"\"\"\nAnalyze all of these together:\n/data/q4_earnings.pdf\n/data/competitor_analysis.pdf\n/data/market_research.xlsx\n/recordings/customer_interview.mp3\n/designs/product_mockup.png\n\nGive me a comprehensive market positioning analysis based on all these inputs.\n\"\"\"\n```\n\nFile paths must be absolute and enclosed in `` tags. CellCog understands PDFs, spreadsheets, images, audio, video, code files, and more.\n\nβ οΈ **Without SHOW_FILE tags, CellCog only sees the path as text β not the file contents.**\n\nβ `Analyze /data/sales.csv` β CellCog can't read the file\nβ
`Analyze /data/sales.csv` β CellCog reads it\n\n### Think of SHOW_FILE like reference files\n\nJust like Nano Banana accepts reference images, CellCog accepts reference files of any type β PDFs, spreadsheets, audio, code, images β as inputs the model reads during a task. Same mental model as any multimodal AI attachment.\n\n**Only attach what you intend to share.** Anything inside a `` tag is uploaded to CellCog. Don't wrap credentials, private keys, `.env` files, SSH keys, or other sensitive material in SHOW_FILE tags β the same way you wouldn't paste them into Nano Banana, ChatGPT, or any other AI service's file upload.\n\n### Request Multiple Outputs, Different Modalities\n\nAsk for completely different output types in ONE request:\n\n```python\nprompt = \"\"\"\nBased on this quarterly sales data:\n/data/sales_q4_2025.csv\n\nCreate ALL of the following:\n1. A PDF executive summary report with charts\n2. An interactive HTML dashboard for the leadership team\n3. A 60-second video presentation for the all-hands meeting\n4. A slide deck for the board presentation\n5. An Excel file with the underlying analysis and projections\n\"\"\"\n```\n\nCellCog handles the entire workflow β analyzing, generating, and delivering all outputs with consistent insights across every format.\n\nβ οΈ **Be explicit about output artifacts.** Without explicit artifact language, CellCog may respond with text analysis instead of generating a file.\n\nβ `\"Quarterly earnings analysis for AAPL\"` β could produce text or any format\nβ
`\"Create a PDF report and an interactive HTML dashboard analyzing AAPL quarterly earnings.\"` β CellCog creates actual deliverables\n\n**Your sub-agent for quality work.** Depth, accuracy, and real deliverables.\n\n---\n\n## Quick Start\n\n### Setup\n\n```python\nfrom cellcog import CellCogClient\n```\n\nIf import fails, install the official CellCog Python SDK:\n```bash\npip install -U cellcog\n```\n\n`cellcog` is the official Python SDK maintained by CellCog AI Inc. Source: https://github.com/CellCog/cellcog_python Β· Package: https://pypi.org/project/cellcog/\n\n### Authentication\n\n**Environment variable (recommended):** Set `CELLCOG_API_KEY` β the SDK picks it up automatically:\n```bash\nexport CELLCOG_API_KEY=\"sk_...\"\n```\n\nGet API key from: https://cellcog.ai/profile?tab=api-keys\n\n```python\nstatus = client.get_account_status()\nprint(status) # {\"configured\": True, \"email\": \"user@example.com\", ...}\n```\n\n### Agent Provider\n\n`agent_provider` is **required** when creating a `CellCogClient`. It identifies which agent framework is calling CellCog β not your individual agent's name, but the platform/tool you're running inside.\n\nExamples: `\"openclaw\"`, `\"claude-code\"`, `\"cursor\"`, `\"aider\"`, `\"windsurf\"`, `\"perplexity\"`, `\"hermes\"`, `\"script\"` (for standalone scripts).\n\n### OpenClaw Agents\n\nFire-and-forget β your agent stays free while CellCog works:\n\n```python\nclient = CellCogClient(agent_provider=\"openclaw\")\nresult = client.create_chat(\n prompt=\"Research quantum computing advances in 2026\",\n notify_session_key=\"agent:main:main\", # OpenClaw session key\n task_label=\"quantum-research\", # Label for notifications\n chat_mode=\"agent\",\n)\n# Returns IMMEDIATELY β daemon delivers results to your session when done\n```\n\n### All Other Agents (Cursor, Claude Code, etc.)\n\nBlocks until done β simplest pattern:\n\n```python\nclient = CellCogClient(agent_provider=\"cursor\") # or \"claude-code\", \"aider\", \"script\", etc.\nresult = client.create_chat(\n prompt=\"Research quantum computing advances in 2026\",\n task_label=\"quantum-research\",\n chat_mode=\"agent\",\n)\n# Blocks until done β result contains everything\nprint(result[\"message\"])\n```\n\n### Credit Usage\n\nCellCog orchestrates 21+ frontier foundation models. Credit consumption is unpredictable and varies by task complexity. Credits used are reported in every completion notification.\n\n---\n\n## Creating Tasks\n\n### Notify on Completion (OpenClaw β Fire-and-Forget)\n\nReturns immediately. A background daemon monitors via WebSocket and delivers results to your session when done. Your agent stays free to take new instructions, start other tasks, or continue working.\n\n```python\nresult = client.create_chat(\n prompt=\"Your task description\",\n notify_session_key=\"agent:main:main\", # Required β your OpenClaw session key\n task_label=\"my-task\", # Label shown in notifications\n chat_mode=\"agent\",\n)\n```\n\n### Wait for Completion (Universal)\n\nBlocks until CellCog finishes. Works with any agent β OpenClaw, Cursor, Claude Code, or any Python environment.\n\n```python\nresult = client.create_chat(\n prompt=\"Your task description\",\n task_label=\"my-task\",\n chat_mode=\"agent\",\n timeout=1800, # 30 min (default). Use 3600 for complex jobs.\n)\nprint(result[\"message\"])\nprint(result[\"status\"]) # \"completed\" | \"timeout\"\n```\n\n### When to Use Which\n\n| Scenario | Best Mode | Why |\n|----------|-----------|-----|\n| OpenClaw + long task + stay free | **Notify** | Agent keeps working, gets notified when done |\n| OpenClaw + chaining steps (research β summarize β PDF) | **Wait** | Each step feeds the next β simpler sequential workflows |\n| OpenClaw + quick task | **Either** | Both return fast for simple tasks |\n| Non-OpenClaw agent | **Wait** | Notify mode is OpenClaw-only |\n\n**Notify mode** is more productive (agent never blocks).\n**Wait mode** is simpler to reason about, but blocks your agent for the duration.\n\n### Continuing a Conversation\n\n```python\n# Wait mode (default)\nresult = client.send_message(\n chat_id=\"abc123\",\n message=\"Focus on hardware advances specifically\",\n)\n\n# Notify mode (OpenClaw)\nresult = client.send_message(\n chat_id=\"abc123\",\n message=\"Focus on hardware advances specifically\",\n notify_session_key=\"agent:main:main\",\n task_label=\"continue-research\",\n)\n```\n\n### Resuming After Timeout\n\nIf `create_chat()` or `wait_for_completion()` times out, CellCog is still working. The timeout response includes recent progress:\n\n```python\ncompletion = client.wait_for_completion(chat_id=\"abc123\", timeout=1800)\n```\n\n### Optional Parameters\n\n```python\nresult = client.create_chat(\n prompt=\"...\",\n task_label=\"...\",\n chat_mode=\"agent\", # See Chat Modes below\n project_id=\"...\", # install project-cog for details\n agent_role_id=\"...\", # install project-cog for details\n enable_cowork=True, # install cowork-cog for details\n cowork_working_directory=\"/Users/...\", # install cowork-cog for details\n)\n```\n\n---\n\n## Response Shape\n\nEvery SDK method returns the same shape:\n\n```python\n{\n \"chat_id\": str, # CellCog chat ID\n \"is_operating\": bool, # True = still working, False = done\n \"status\": str, # \"completed\" | \"tracking\" | \"timeout\" | \"operating\"\n \"message\": str, # THE printable message β always print this in full\n}\n```\n\n**β οΈ Always print the entire `result[\"message\"]`.** Truncating or summarizing it will lose critical information including generated file paths, credits used, and follow-up instructions.\n\n### Utility Methods\n\n**`get_history(chat_id)`** β Full chat history (when original delivery was missed or you need to review). Returns the same shape; if still operating, `message` shows progress so far.\n\n```python\nresult = client.get_history(chat_id=\"abc123\")\n```\n\n**`get_status(chat_id)`** β Lightweight status check (no history fetch):\n\n```python\nstatus = client.get_status(chat_id=\"abc123\")\nprint(status[\"is_operating\"]) # True/False\n```\n\n---\n\n## Chat Modes\n\n| Mode | Best For | Speed | Min Credits |\n|------|----------|-------|-------------|\n| `\"agent\"` | Most tasks β images, audio, dashboards, spreadsheets, presentations | Fast (seconds to minutes) | 100 |\n| `\"agent core\"` | Coding, co-work, terminal operations | Fast | 50 |\n| `\"agent team\"` | Deep research & multi-angled reasoning across every modality | Slower (5-60 min) | 500 |\n| `\"agent team max\"` | High-stakes work where extra reasoning depth justifies the cost | Slowest | 2,000 |\n\n- **`\"agent\"` (default)** β Most versatile. Handles most tasks excellently, including deep research when guided.\n- **`\"agent core\"`** β Lightweight context for code, terminal, and file operations. Multimedia tools load on demand. Requires Co-work (CellCog Desktop). See `code-cog`.\n- **`\"agent team\"`** β A team of agents that debates, cross-validates, and delivers comprehensive results. The only platform with deep reasoning across every modality.\n- **`\"agent team max\"`** β Same Agent Team with all settings maxed. Quality gain is incremental (5-10%) but meaningful for costly decisions.\n\n---\n\n## Working with Files\n\n### Input: SHOW_FILE\n\nInclude local file paths in your prompt with `` tags (absolute paths required):\n\n```python\nprompt = \"\"\"\nAnalyze this sales data and create a report:\n/path/to/sales.csv\n\"\"\"\n```\n\n### Output: GENERATE_FILE\n\nUse `` tags to specify where output files should be stored on your machine. Essential for deterministic workflows where the next step needs to know the file path in advance.\n\n```python\nprompt = \"\"\"\nCreate a PDF report on Q4 earnings:\n/workspace/reports/q4_analysis.pdf\n\"\"\"\n```\n\nOutput downloads to the specified path instead of default `~/.cellcog/chats/{chat_id}/`.\n\n### File Downloads\n\nThe SDK automatically downloads files from CellCog responses:\n- **If you used `GENERATE_FILE` tags:** Files download to the path you specified\n- **Otherwise:** Files download to `~/.cellcog/chats/{chat_id}/`\n\nDownloaded file paths appear in `result[\"message\"]`. The SDK tracks seen messages β files are only downloaded once.\n\n**If you missed files or need to re-sync:**\n```python\nresult = client.get_history(chat_id=\"abc123\")\n```\n`get_history()` re-processes the entire chat and downloads any missed files to their original destinations.\n\n---\n\n## Tips\n\n### β οΈ CellCog Web Fallback\n\nEvery chat is accessible at https://cellcog.ai. When work gets complex or the SDK hits issues, direct your human to the web platform to view, continue, or take over directly.\n\n---\n\n## What CellCog Can Do\n\nCellCog is a sub-agent β not an API. Your agent offloads complex work to CellCog, which reasons, plans, and executes multi-tool workflows internally. A proprietary agent-to-agent communication protocol ensures high accuracy on first output, and because these are agent threads (not stateless API calls), every aspect of every generation can be refined through multi-step iteration.\n\nUnder the hood: frontier models across every domain, upgraded weekly. CellCog routes to the right models automatically β your agent just describes what it needs.\n\nInstall capability skills for detailed guidance:\n\n| Category | Skills |\n|----------|--------|\n| **Research & Analysis** | `research-cog` `fin-cog` `crypto-cog` `data-cog` `news-cog` |\n| **Video & Cinema** | `video-cog` `cine-cog` `insta-cog` `tube-cog` `seedance-cog` |\n| **Images & Design** | `image-cog` `brand-cog` `meme-cog` `banana-cog` `3d-cog` `gif-cog` `sticker-cog` |\n| **Audio & Music** | `audio-cog` `music-cog` `pod-cog` |\n| **Avatars & Personas** | `avatar-cog` |\n| **Documents & Slides** | `docs-cog` `slides-cog` `spreadsheets-cog` `resume-cog` `legal-cog` |\n| **Apps & Prototypes** | `dash-cog` `game-cog` `proto-cog` `diagram-cog` |\n| **Creative** | `comi-cog` `story-cog` `learn-cog` `travel-cog` |\n| **Development** | `code-cog` `cowork-cog` `project-cog` `think-cog` |\n\n**This skill shows you HOW to use CellCog. Capability skills show you WHAT's possible.**\n\n---\n\n## OpenClaw Reference\n\n### Session Keys\n\nThe `notify_session_key` tells CellCog where to deliver results:\n\n| Context | Session Key |\n|---------|-------------|\n| Main agent | `\"agent:main:main\"` |\n| Sub-agent | `\"agent:main:subagent:{uuid}\"` |\n| Telegram DM | `\"agent:main:telegram:dm:{id}\"` |\n| Discord group | `\"agent:main:discord:group:{id}\"` |\n\n**Resilient delivery:** If your session ends before completion, results are automatically delivered to the parent session (e.g., sub-agent β main agent).\n\n### Sending Messages During Processing\n\nIn notify mode, your agent is free β you can send additional instructions to an operating chat at any time:\n\n```python\nclient.send_message(chat_id=\"abc123\", message=\"Actually focus only on Q4 data\",\n notify_session_key=\"agent:main:main\", task_label=\"refine\")\n\nclient.send_message(chat_id=\"abc123\", message=\"Stop operation\",\n notify_session_key=\"agent:main:main\", task_label=\"cancel\")\n```\n\nIn wait mode, your agent is blocked and cannot send messages until the current call returns.\n\n---\n\n## Support & Troubleshooting\n\nFor error handling, recovery patterns, ticket submission, and daemon troubleshooting:\n\n```python\ndocs = client.get_support_docs()\n```\n","tags":{"latest":"2.0.15"},"stats":{"comments":2,"downloads":14393,"installsAllTime":103,"installsCurrent":103,"stars":8,"versions":43},"createdAt":1770104190626,"updatedAt":1778988017357},"latestVersion":{"version":"2.0.15","createdAt":1776927732615,"changelog":"- Documentation clarifies that anything wrapped in tags is uploaded to CellCog; warns not to include credentials, private keys, or sensitive material as SHOW_FILE files.\n- Added a comparison to reference image attachments in Nano Banana to explain usage.\n- Installation section now specifies that cellcog is the official Python SDK, with source and PyPI links.\n- Wait/Notify mode comparison table updated: Notify mode is now OpenClaw-only, not available for other agents; clarified table language and guidance.\n- Minor edits for accuracy and improved guidance throughout the setup and usage sections; small formatting improvements.","license":"MIT-0"},"metadata":{"setup":[{"key":"CELLCOG_API_KEY","required":true}],"os":["darwin","linux","windows"],"systems":null},"owner":{"handle":"nitishgargiitd","userId":"s1766yx1dct3dffeec508gjrz983fce7","displayName":"CellCog","image":"https://avatars.githubusercontent.com/u/14193431?v=4"},"moderation":null}