Skip to content

Rapidly MCP

Test the idea before you build it. Rapidly works inside your AI agent: in about five minutes it takes an idea through a Lean Canvas, turns the riskiest assumption into a measurable hypothesis with a researched pass mark, and returns a build prompt for the smallest Pretotyping experiment worth running. The agent builds the test; real customers provide the evidence. Work is saved to your Rapidly team instead of disappearing in a chat thread.

Server type: Remote Streamable HTTP (hosted)
Endpoint: https://www.rapidly.co/mcp
Auth: one-time bearer token (RAPIDLY_MCP_TOKEN) for new teams, OAuth for existing teams
Tools: 21 (idea generation, Lean Canvas design, hypothesis and experiment design, evidence interpretation, portfolio scoring, team records)
Repo: github.com/Exponentially-Platform/Rapidly-MCP (MIT, 1 star)
Registry: co.rapidly/rapidly
Pricing: free trial with five new ideas; team plans by request

Why This Matters for Operators

Asking a general model for a business plan produces something that looks reasonable and proves nothing. Rapidly encodes the method instead: Alberto Savoia's Pretotyping combined with the vendor's practice across 4,000+ client experiments and 200+ workshops. The core run builds a full Lean Canvas, converts the riskiest demand assumption into a market-engagement hypothesis with a researched pass threshold, then designs three to five named experiments and recommends the smallest useful one with a build prompt. The output includes GO/NO-GO rules, cited sources, and a test asset your agent can build immediately.

Tools & Capabilities

Tool What it does
independent_generate_ideas Generates grounded ideas from a company, problem, or brief with cited public sources
independent_design_lean_canvas Builds a complete Lean Canvas, preserving the original customer, problem, and proposition
independent_design_hypothesis Turns the riskiest assumption into a market engagement hypothesis with a researched pass threshold
independent_design_experiments Creates three to five named Pretotyping experiments, recommends the smallest useful one, returns its build prompt
design_hypothesis Designs a hypothesis and researched target from a saved idea or raw idea text
design_experiments Designs three to five Pretotyping options for a saved idea
interpret_results Turns recorded experiment evidence into an executive summary
evaluate_idea Summarises the evidence and recommends what to do next
score_portfolio Orders the team's ideas by stored value and effort, with confidence and experiment counts
get_sprint_pack Combines the ranked portfolio, team knowledge, focus idea, hypothesis, and fresh experiment designs
list_ideas Lists the team's idea portfolio
get_idea Opens one idea and its current method work
list_experiments Lists an idea's experiments, outcomes, and lessons
get_team_knowledge Reads company context and lessons saved by the team
save_idea Updates selected fields on an existing idea
expand_idea Stages proposed Canvas, hypothesis, and experiment work for review
save_experiment Saves a new experiment with the idea it tests
save_team_knowledge Adds company context or a lesson to team knowledge
create_idea_bundle Creates an idea with optional Canvas, hypothesis, and experiment records
generate_ideas Generates grounded, de-duplicated ideas inside an existing team portfolio
validate_idea Produces a quick Canvas, hypotheses, and a suggested first experiment without public-source grounding

Installation

Open rapidly.co/rapidly-mcp, enter a name, email, and team name. Rapidly creates the account and team, then shows the personal token once (no plan or password needed). The page provides copy-ready setup for Codex CLI and Claude Code.

Configuration

  • New teams: keep the bearer token in the RAPIDLY_MCP_TOKEN environment variable and register the endpoint https://www.rapidly.co/mcp with that header.
  • Existing teams: connect with OAuth, choose the team, and approve access.
  • Works with Claude, Codex, ChatGPT, and Hermes; any agent that supports remote MCP can connect.
  • Rapidly checks team and access scoping on every request.

Business Relevance

  • Pre-build validation: a build prompt for the smallest experiment instead of a full product commitment.
  • Evidence-based decisions: GO/NO-GO rules and pass thresholds reviewed by real customers.
  • Portfolio discipline: score_portfolio ranks ideas by value and effort for funding conversations.
  • Institutional memory: experiments, outcomes, and lessons stay in the team workspace.

Integration with CorpusIQ

Rapidly's experiment evidence pairs with CorpusIQ connectors for the measurement side: validate demand with GA4 landing-page traffic, Shopify trial orders, or Stripe revenue during a Fake Door test, and log hypotheses and pass marks in Notion or Airtable so the validation record lives next to the operating data.

Limitations

  • Free trial caps at five new ideas; team plans require contacting the vendor.
  • Generated facts, cited sources, and proposed targets need human review before use.
  • Bearer token is shown once at creation - store it securely.
  • Young repo (created Aug 24, 2026, 1 star); anonymous endpoint probes return 401 (liveness verified).

See Also

Contact Terms Privacy Pricing