Chiron’s Forge runs your request through multiple AI research engines, synthesizes the findings, and has an independent model judge the result — refining until it passes. You get a finished skill file, research report, or Cursor ruleset. Not a draft. A deliverable.
Your first build is free. No card. No subscription — you pay once, per build.
Every build starts the same way — you describe what you need, and the research pipeline goes to work. What changes is the format you get at the end. Pick the one your team will actually use.
A single, procedural, action-oriented skill file — the kind you drop into Claude, an agent framework, or any tool that reads SKILL.md. It’s scored by an independent judge before you get it, so it behaves like domain knowledge, not a guess.
A focused 4–6 page report: executive summary, findings by theme, recommendations, and a source list with real citations. For medical and scientific work, reports can be evidence-graded — drawing on PubMed, Semantic Scholar, and OpenAlex, with every claim tied to a DOI and the grade verified before delivery.
A .cursor/rules bundle that drops straight into your project. The free version is built from the stack you declare. The paid version reads your actual repository — your naming, your structure, your patterns — and writes rules no template could reproduce.
Most AI tools hand you the first thing the model produced. Chiron’s Forge doesn’t. Here’s what runs before anything reaches you.
One model writes. A different model — never the same one, never Anthropic — scores the result against a rubric. The build keeps refining until it clears the bar, and the best-scoring version is the one you get. Not the last one. The best one.
Depending on your tier, Gemini, OpenAI, and DeepSeek research your topic at the same time. Their findings are synthesized and gap-checked — so the output isn’t one model’s blind spots repeated, it’s several independent passes reconciled.
Between research passes, the pipeline checks what’s missing and goes back for it — up to four times on the deepest tier. A weak answer triggers more research, not a shrug.
Upload CSV, Excel, or PDF files to ground a build in your own material. Personal information is redacted before analysis, and the files are deleted when the build ends. Nothing is kept.
Chat with Claude to scope exactly what you need.
Optionally attach your own data to ground the build.
Multiple AI engines research your topic in parallel.
An independent model scores the result and sends it back for refinement until it passes.
Get your skill file, report, or ruleset — finished, not a first draft.
You’re not renting access. Each build is a one-time purchase, and your first one is free. Pick your output, pick your depth.
A SKILL.md your agents load at runtime.
SKILL.md · judge-scored · plain Markdown you own and export
Your first build, no card. One research pass, synthesized, SKILL.md download.
Try freeTwo research engines, gap-checked, judge-scored (75/100 bar), one refinement pass.
Start a buildDeep research mode, three passes, two refinement passes, data upload.
Start a buildA cited brief you can hand to a decision-maker.
.zip (report.md + sources.json) · citations included · evidence grading available
Two engines, standard depth, cited report.
Start a buildDeep research mode, three passes, evidence grading available.
Start a buildMaximum depth — all three engines, four passes, highest judge bar, full evidence grading.
Start a buildAcademic sources included: Pro and Signature reports add PubMed, OpenAlex, and Semantic Scholar grading at no extra cost. Signature unlocks clinical-grade evidence (Bronze / Silver / Gold).
A ruleset that matches how your codebase actually works.
.cursor/rules bundle · 4–6 .mdc files · repo-aware on Signature
Rules built from the stack you declare.
Try freeRepo-aware rules — Chiron's Forge reads your actual codebase and writes to its conventions.
Start a buildAll prices in USD · one-time, per build · your first build is free.
Describe what you need. See what comes back. Decide from there.
No card required. No subscription. One free build to see the quality for yourself.