%20scale(1.25)'%20fill='none'%20stroke='hsl(224%2064%25%2033%25)'%20stroke-width='1.8'%20stroke-linecap='round'%20stroke-linejoin='round'%3e%3cpath%20d='M9%204H5v16h4M15%204h4v16h-4'/%3e%3ccircle%20cx='12'%20cy='12'%20r='2.5'%20fill='hsl(224%2064%25%2033%25)'%20stroke='none'/%3e%3c/g%3e%3cg%20transform='translate(40%2028.5)%20scale(0.024%20-0.024)'%20fill='hsl(224%2040%25%2012%25)'%3e%3cpath%20transform='translate(0)'%20d='M79%200V710H213V371L152%20415H567L506%20371V710H640V0H506V343L567%20299H152L213%20343V0Z'/%3e%3cpath%20transform='translate(664)'%20d='M108%20-150V-48H177Q203%20-48%20215.5%20-40.5Q228%20-33%20235%20-14L249%2022H211L21%20534H158L290%20146L415%20534H552L344%20-55Q326%20-106%20293%20-128Q260%20-150%20201%20-150Z'/%3e%3cpath%20transform='translate(1170)'%20d='M69%20-150V534H196L199%20421L187%20426Q207%20485%20252.5%20515.5Q298%20546%20358%20546Q434%20546%20484.5%20508.5Q535%20471%20560%20408Q585%20345%20585%20267Q585%20190%20559.5%20126.5Q534%2063%20483.5%2025.5Q433%20-12%20357%20-12Q318%20-12%20283.5%202Q249%2016%20224.5%2042Q200%2068%20189%20104L202%20113V-150ZM325%2094Q382%2094%20415%20140Q448%20186%20448%20267Q448%20348%20415%20394Q382%20440%20325%20440Q287%20440%20259.5%20421Q232%20402%20217%20363Q202%20324%20202%20267Q202%20210%20216.5%20171.5Q231%20133%20259%20113.5Q287%2094%20325%2094Z'/%3e%3cpath%20transform='translate(1739)'%20d='M303%20-12Q224%20-12%20164%2022.5Q104%2057%2071.5%20120Q39%20183%2039%20267Q39%20352%2071.5%20414.5Q104%20477%20164%20511.5Q224%20546%20303%20546Q383%20546%20442.5%20511.5Q502%20477%20534.5%20414.5Q567%20352%20567%20267Q567%20183%20534.5%20120Q502%2057%20442.5%2022.5Q383%20-12%20303%20-12ZM303%2094Q364%2094%20397%20139.5Q430%20185%20430%20267Q430%20349%20397%20394.5Q364%20440%20303%20440Q243%20440%20209.5%20394.5Q176%20349%20176%20267Q176%20185%20209.5%20139.5Q243%2094%20303%2094Z'/%3e%3cpath%20transform='translate(2273)'%20d='M285%200Q204%200%20166.5%2037Q129%2074%20129%20154V660H261V168Q261%20130%20277%20116Q293%20102%20327%20102H402V0ZM44%20433V534H402V433Z'/%3e%3cpath%20transform='translate(2647)'%20d='M69%200V534H202V0ZM67%20605V719H204V605Z'/%3e%3cpath%20transform='translate(2863)'%20d='M50%200V109L356%20433H62V534H513V426L203%20102H520V0Z'/%3e%3cpath%20transform='translate(3358)'%20d='M302%20-12Q222%20-12%20162.5%2022.5Q103%2057%2071%20120Q39%20183%2039%20267Q39%20352%2071%20414.5Q103%20477%20162%20511.5Q221%20546%20300%20546Q377%20546%20434.5%20512Q492%20478%20523.5%20413.5Q555%20349%20555%20260V231H176Q180%20161%20213.5%20127Q247%2093%20304%2093Q346%2093%20373.5%20111.5Q401%20130%20412%20164L547%20156Q526%2078%20460.5%2033Q395%20-12%20302%20-12ZM176%20318H418Q414%20382%20382%20413Q350%20444%20300%20444Q249%20444%20216%20411.5Q183%20379%20176%20318Z'/%3e%3c/g%3e%3c/svg%3e)
Skills & plugin
One install. Your agent runs growth experiments the Hypotize way.
The Hypotize plugin brings the MCP server and three skills to Claude Code and Codex. The server enforces the rules. The skills teach your agent to work inside them: read before proposing, score with evidence, close every test with a measured result.
Install
Pick your agent, paste two lines
No API key. The first time your agent uses Hypotize, it opens a sign-in page and you choose what it may do. You can revoke it from your project's Connections page at any time.
Claude Code
Adds the marketplace, then installs the MCP server and the three skills.
/plugin marketplace add viluca94/hypotize-plugin/plugin install hypotizeCodex
Same plugin, same skills, from the Codex CLI.
codex plugin marketplace add viluca94/hypotize-plugincodex plugin add hypotize@hypotizeCursor, ChatGPT and other MCP clients
Connect guideAdd the server URL as a remote MCP server. The connect guide has the steps for each client.
https://api.hypotize.ai/mcpWhat is inside
Three skills on top of 21 tools
The MCP server gives the agent 21 tools on your experiment record. Each skill is a plain SKILL.md file the agent loads when the task matches, so it hits fewer refusals and leaves a cleaner record.
The method
hypotize-growth-experiments
The agent reads the backlog and past learnings before it proposes anything, writes the hypothesis first, votes in the experiment's framework with evidence, follows the stages in order and closes a test only with a measured result.
Use it for: Proposing, scoring, starting, updating or closing an experiment.
The weekly review
hypotize-weekly-review
A read-only pass over the pipeline: finished tests waiting for a verdict, tests running past their due date, committed work nobody started. It ends with the decisions the team has to take, and moves nothing on its own.
Use it for: Stand-up prep, a pipeline review, "where are we on our tests".
The backlog import
hypotize-backlog-import
The agent reads the whole source, skips what is already in Hypotize, shows you the field and status mapping before it writes, then imports in atomic batches of 50.
Use it for: Moving a spreadsheet, a Notion database or a doc into Hypotize.
Skills without the plugin
Already connected through MCP?
Copy the skill folders into your Claude Code skills directory. In Claude.ai or Claude Desktop, upload the SKILL.md files as project knowledge. Any other client can use a SKILL.md body as its instructions for Hypotize work.
git clone https://github.com/viluca94/hypotize-plugin.gitcp -r hypotize-plugin/skills/hypotize-* ~/.claude/skills/Open source
Read every line before you install it
The plugin is MIT licensed and public on GitHub: the manifests, the server URL and the three skills. The connect reference lists every tool the agent gets and what it cannot do.