Stardock's desktop app for running persistent, named AI staff that orchestrate command-line agents like Claude Code and Gemini from one workspace. Genuinely capable, on the condition you become its administrator first.
Marketers and operators who already work in Claude Code or other command-line agents and want a persistent, named staff layer with shared workspace memory.
You want every AI tool in one window. Clairvoyance does not reach Perplexity, ChatGPT web, or Gemini web; it orchestrates command-line agents only.
A real persistent-staff orchestration layer for CLI power users. Strong staff behaviour and a steep setup, improving fast across its beta releases, with one output-truncation flaw to manage.
A workspace of persistent, named AI staff that delegate to temporary specialists, all running on command-line agents you already pay for. The orchestration is between you and your staff, not a black box that calls out to other vendors on its own.
Scored from a multi-week hands-on field test, retested on the Beta 2 / v0.66 build. Performance is held down by a research-output truncation issue; everything else improved markedly across the beta.
AI workers with their own names, roles, and memory that live in your workspace.
Temporary specialists you spawn for a single task; they finish, report, and dismiss.
Drives Claude Code, Codex, Gemini, Copilot, and Cursor, plus local Ollama models.
A shared directory of documents and notes that staff read and write natively.
Exposes its own tools, so you can drive Clairvoyance from another agent session.
Built-in task tracking with dependencies, sprint planning, and scheduled runs.
The app is free in beta; the real cost is whatever model usage you already pay for. Pricing shown as of the review date and subject to change.
We ran Clairvoyance as a multi-week practitioner field test, using it for real editorial research and drafting rather than a staged demo. That included setting up a workspace, importing our house style guide into staff memory, delegating live research to spawned specialists, and returning after a long break to see exactly what the staff remembered. We then retested on a newer beta that had shipped in the meantime, which changed several of our findings for the better. The notes from both passes, friction and wins alike, are what this review is built on. Clairvoyance is a fast-moving beta in active development by Brad Wardell and the Stardock team, so we have judged it as a work in progress and flagged the rough edges that look most likely to improve.
I came back to Clairvoyance after a few weeks away and a couple of version jumps, and it had changed more than I expected. On the Beta 2 build, most of what frustrated me early had been fixed or softened. The memory I thought it had lost turned out to be recoverable; the cost picture, once buried, is now a clear usage card running off my existing subscription; and it runs on Windows, Mac, and Linux, not just one platform.
What stayed true is the character of the thing: capable, honest staff that verify before they answer, wrapped in a setup that asks real effort from you before it pays off. One genuine flaw remains, a hidden ceiling that quietly cuts off the tail of long results, and that is the main reason I am not scoring it higher. For a marketing or e-commerce practitioner who already works in command-line agents and will invest the setup time, Clairvoyance earns a place. For everyone else, it is one to watch closely, because it is improving fast.
MetaGem
Note: factual fields above are filled before scoring; this verdict and the scores are added only after real hands-on use.
The app is a free download, currently in beta. The models it runs still bill to your own accounts, so the real cost is whatever Anthropic, Google, or OpenAI usage you already pay for. Local models through Ollama add no per-token cost.
No. Clairvoyance orchestrates command-line agents, not web subscriptions. It does not reach Perplexity Pro, ChatGPT web, or Gemini web. If your goal is one window for all your AI subscriptions, this is not that tool.
Yes, with a caveat. Your conversation history is preserved and recoverable with a resume command. What is not automatic is the staff volunteering that context after a cold restart; its own working notes and the resumable history are separate, so you resume deliberately. Keeping your canonical record outside the tool is still wise.
Command-line agents including Claude Code, OpenAI Codex, Gemini, GitHub Copilot, and Cursor, plus local models through Ollama.
Not yet, in our testing. The memory and output limits make us cautious about anything that needs to run reliably while you are not watching. For scheduled operations, a dedicated automation platform is still the safer choice.
Clairvoyance is a real and interesting orchestration layer for people who already live in command-line agents, and a recent beta fixed much of what I first criticised. Its staff are honest and capable; the one habit to watch is a hidden ceiling that silently truncates long results passed between staff. Worth a weekend if you fit the profile, and worth watching whether you do or not.
| Clairvoyance | Claude Code (raw) | n8n | |
|---|---|---|---|
| Starting price | Free (Beta) | Free CLI | Free (self-host) |
| Best for | CLI staff & delegation | Single-agent CLI work | Scheduled automations |
| Learning curve | Steep | Moderate | Moderate |
| Persistent memory | Resumable sessions | Your own files | External store |
| MetaGem Score | 7.0 | Not yet tested | Not yet tested |
Field-tested AI workflows, honest tool notes, and MetaGem updates for marketers, business owners, and e-commerce teams. No hype, just what works.
No spam. Unsubscribe anytime.
