TriForge AI Code Council runs OpenAI, Claude, and Grok as a structured deliberation panel directly in your editor. Draft, critique, debate, and ship code with three AI minds reaching consensus — without leaving VS Code.
✓ 1-day free trial — no credit card required · Solo mode always free
Every AI coding tool gives you one answer from one model. TriForge AI Code Council is different. It runs three frontier AI models simultaneously — OpenAI GPT, Anthropic Claude, and xAI Grok — and puts them through a structured deliberation pipeline before showing you the result.
First, one model drafts. Then all three critique for bugs, security issues, and logical gaps. Then they debate. Then they reach consensus. What you get isn't a suggestion — it's a vetted solution.
Solo mode works with any single provider and is always free. Full Council mode — with cross-critique and consensus — requires a subscription after the 1-day trial.
Every Council run follows the same structured deliberation pipeline — the same way a senior engineering team reviews code before it ships.
The fastest available model generates an initial implementation. This is the raw first draft — optimized for speed, not perfection. It becomes the subject of scrutiny in the phases that follow.
TriForge scans the draft for security vulnerabilities, injection risks, logic errors, and edge cases. The risk level (LOW / MEDIUM / HIGH / CRITICAL) adjusts how intensely the council debates in the next phase.
Each model reviews the draft independently and delivers a structured verdict — agree, disagree, or partial. Models with objections must explain the specific flaw and propose a concrete fix. No vague rejections.
When models disagree, the council enters debate mode. Models challenge each other's reasoning directly. TriForge mediates — it can call synthesis, extended debate, or escalate to you for a decision.
The council reaches a final verdict. The agreed solution is shown alongside a confidence score. You can apply it directly to your editor, preview it in VS Code's diff viewer first, or request an alternative take from any model.
Code Council is a native VS Code extension. Every feature lives inside the editor — no browser, no context switching.
OpenAI, Claude, and Grok deliberate in parallel. Each model plays a structured role — drafter, critic, or synthesizer — depending on the phase.
Select any code and right-click to Explain, Write Tests, Refactor, or Find Bugs. The council opens automatically with your selection pre-loaded.
Stage, unstage, commit, push — all from the Council panel. See a diff preview of staged changes, browse recent commits, and manage branches without leaving the extension.
Browse your project file tree and attach files as context for any council run. Add specific files to ground the AI's analysis in your actual codebase.
When the council produces a solution, review it side-by-side in VS Code's native diff editor before accepting. One click to apply, one to discard.
Every prompt you run is saved in a searchable history panel. Re-run any previous query, or use it as a starting point for a new one.
Choose which specific model each provider uses directly from the panel — GPT-4o, Claude Sonnet, Grok-3, and more. No settings file required.
Set the council's debate intensity from Cooperative to Ruthless. Higher intensity means more adversarial critique — useful for security-critical or production-bound code.
Export the full deliberation — all model responses, risk analysis, debate rounds, and final consensus — as a formatted Markdown file for documentation or review.
Use your own API keys. You pay OpenAI, Anthropic, and xAI directly — TriForge never marks up your API usage. The subscription covers access to the council orchestration engine.
1 day free · then $15/month · cancel anytime
TriForge AI for Mac and Windows brings the same three-model think tank to your entire desktop — with voice input, long-term memory, file system access, and more. Free to start, Pro and Business plans available.