How do I get a PAT?
Free for all Colosseum Arena members. Log in to Colosseum Arena and generate a token. Set two environment variables:Which tools are supported?
| Tool | Status | Install method |
|---|---|---|
| Claude Code | Supported | npx skills add ColosseumOrg/colosseum-copilot |
| Codex | Supported | npx skills add ColosseumOrg/colosseum-copilot -a codex |
| OpenClaw | Supported | npx skills add ColosseumOrg/colosseum-copilot -a openclaw |
What are the rate limits?
See API Reference for the full rate limit table.What happens when my idea already exists?
Copilot tells you directly. It runs incumbent validation and classifies the gap:- Full gap: nobody has addressed this problem
- Partial gap: incomplete coverage (segment, UX, geographic, pricing, or integration)
- False gap: already solved
Is my research data private?
Yes. Query content and results are never stored, shared, or used for training. We log request metadata (route, status code, latency) for reliability monitoring, but your actual queries and responses are not persisted.How current is the data?
- Hackathon projects update after each Colosseum hackathon
- Archive sources refresh on varying intervals (protocol docs within days, historical archives are static)
- The Grid provides crypto ecosystem metadata
- Web search is live
- Breakpoint transcripts are added after each annual conference
What’s the difference between conversational and deep dive mode?
Conversational (default): Answers questions with targeted API calls and inline citations. Fast, concise, evidence-backed. Good for lookups, comparisons, and quick assessments. Deep dive (explicit opt-in): Full 8-step research workflow with parallel data gathering, incumbent validation, gap classification, and a structured report including revenue model, GTM strategy, and risk assessment. Activates when you say “vet this idea”, “deep dive”, or “should I build X?”Can I use the API directly without the skill?
Yes. The skill file is a prompt that teaches your coding assistant how to use the API effectively, but the underlying REST endpoints work with any HTTP client. See API Reference for curl examples.What if my query returns no results?
Archive search auto-cascades through three retrieval tiers (vector → chunk text → document text) before returning empty. If you still get no results:- Try conceptual synonyms (e.g.,
"prediction markets"→"futarchy") - Broaden the query (3–6 keywords work best)
- Remove filters that may be too restrictive
- Check if the topic exists in the corpus (some niche areas may not have coverage yet)
Does Copilot write code?
No. Copilot is a research tool, not a code generator. It helps you understand the landscape — what’s been built, who’s funded, where the gaps are — so you build the right thing. It runs inside your coding assistant but doesn’t produce Solana programs or smart contracts.Does Copilot replace doing my own research?
No. Think of it as a starting point that covers ground faster than you can manually. Copilot searches structured datasets and curated archives, but you should always verify the current status of projects and market data independently. Hackathon submissions are demos and prototypes — not all become active products.How is this different from asking ChatGPT about Solana?
Data access. General-purpose LLMs don’t have structured access to 5,400+ Colosseum hackathon submissions, 84,000+ curated archive documents, or The Grid’s ecosystem metadata. Copilot’s evidence floors also mean every claim must trace to a specific source — it won’t fill gaps with speculation.How do I report issues or suggest sources?
Use thePOST /feedback endpoint — your agent can call this automatically when it encounters issues. You can also reach out in the Colosseum Discord.
To suggest a new archive source, use POST /source-suggestions or see Archive Corpus: Source freshness and suggestions.