Run QuReDec from Claude Code, Cursor, or any script

For most of QuReDec™'s first month, the only way to generate a decision brief was to open quredec.com/app in a browser, type your question into the Decision Coach, and wait two to five minutes. That works — and most of our users will keep doing exactly that. But a decision brief is a piece of structured data. If you're already living inside Claude Code, Claude Desktop, Cursor, or your own scripts, opening a browser tab to get one feels like the wrong shape.

So we shipped two new surfaces. Both are live now on every QuReDec account.

The REST API

You mint a per-user API key from your account page, authenticate with a Bearer token, POST /api/v1/brief with a question, and poll for the result. The brief is the same one the web app produces — same recommendation, same confidence score, same cited sources, same shareable URL.

curl -X POST https://quredec.com/api/v1/brief \
  -H "Authorization: Bearer qrd_live_..." \
  -H "Content-Type: application/json" \
  -d '{"question":"Should we adopt parametric release for our sterilization line?"}'

The response includes a brief_run_id you can poll, and a public_share_url you can hand to a colleague. One credit per brief — same as the web app. Full reference is at quredec.com/docs/api.

The MCP server

The other surface is a Model Context Protocol server. MCP is the spec Anthropic introduced for AI assistants to talk to external tools — Claude Desktop, Claude Code, and Cursor all support it. We packaged the QuReDec API into a tiny Python server, quredec-mcp, that runs on your machine and exposes three tools:

• decision_brief — submit a question and (by default) wait for the brief.
• get_brief_status — poll a previously submitted brief.
• list_recent_briefs — see what's been generated under this account.

Once it's wired into your AI client's config, you can ask the assistant a research-grade question and the brief renders inside the conversation:

"Use the decision_brief tool to evaluate whether we should migrate from Postgres to a managed cloud database, given a $200K/yr spend and 4 engineers."

The assistant calls the tool, the QuReDec engine pulls evidence from Brave Search, Semantic Scholar, arXiv, Crossref, and the rest of the source mix, synthesizes the brief, and returns it inline with the recommendation, key facts (with citations), implications, recommended actions, risks, and a public share URL — all without leaving the conversation. The PDF is downloadable from the share URL once you sign in.

Setup is in the docs; the server publishes to PyPI as quredec-mcp.

Why this matters

Two reasons. First, decisions don't happen in a browser tab — they happen in the middle of writing code, reviewing a contract, drafting a memo. Reaching the brief from inside the surface where the work is already happening makes the cost of asking near-zero, and that changes how often a decision actually gets researched instead of guessed.

Second, the brief is now composable. You can chain it into a script, a cron job, or a longer agentic workflow. The output is structured JSON with stable fields (recommendation, confidence, key_facts[].citation_ids, citations[]) — so it's a real input to other systems, not just a thing a person reads.

How to get started

Sign in at quredec.com/account, scroll to API Keys, click Generate new key, and copy the token shown once. The full reference (curl, Python, Node, plus Claude Desktop / Claude Code / Cursor configs) is at quredec.com/docs/api. If you hit anything weird, email us and we'll dig in.

Same engine. Same citations. Same evidence-backed brief. Just reachable from wherever you already work.