StateWeave runs your AI agent in sandboxes of the apps it works in: stateful clones of Slack, Gmail, Salesforce, Shopify, Zendesk, Google Calendar, Sheets, Drive, WhatsApp and Teams, with the same APIs, MCP servers and UIs as the real thing. Describe a scenario to fill every app with consistent data, inject faults like outages and 401s, and get graded runs showing which checks failed and the exact calls behind them. Fresh copy every run.
StateWeave helps teams test AI agents on working copies of the apps they actually use, such as Slack, Gmail, Salesforce, Shopify and Zendesk and more.
Instead of relying on mocked APIs that only return what you wrote, StateWeave lets your agent run against twins with the same APIs, MCP servers, web UIs and errors as the real apps, filled with a realistic scenario, so you can see what it gets wrong before it ships.
What stands out:
- Working copies of real applications, with your agent’s existing client libraries unchanged
- Custom app twins on demand: tell the team which app you need, and they'll build a working replica for you
- Generate a whole consistent company or scenarios (people, history, records) from one sentence, or bring your own data.
- Inject real-world failures like outages, expired tokens and cut-off responses
- Grade every run and compare agent versions side by side
- Let Claude Code, Codex or Cursor run the tests, read what broke and fix it
mocked apis lie, stateful clones dont. this is so usefull for agent evals 👏
Gmail for me — my agents touch real accounts, and the nightmare scenario is one sending or archiving something real while I'm looking the other way. The fresh-copy-per-run thing plus opening the replica to see exactly what it did is the part I'd actually lean on, since 'it said it worked' never cuts it when the blast radius is a real inbox.
This is a real pain point. We test our agents against our production workspace more often than we'd like to admit. How close are the clones to the real APIs, for example rate limits and permission errors?
testing agents on salesforce clones instead of live deals, smart move. congrats Mike!
salesforce for sure, that closed lost story is too real lol. can the fault injection do a write that times out but actually went thru? thats the one that makes our agents create dupes
Hi Product Hunt 👋 I'm Jinbiao (Mike), founder of StateWeave!
I built this because testing agents against real apps kept leaving a mess. Once, a test run in the real Salesforce marked a live deal as Closed Lost. We found out the next morning when the pipeline report dropped, and we spent the day rebuilding the record by hand. There’s no reset button on a real workspace.
Most teams building AI agents (startups, FDEs, platform teams) run into the same problem. They can’t safely test against the real apps their agents touch, so they ship and find the problems in production, or test on stateless mocks.
StateWeave gives your agent a sandbox with working replicas of those apps: Slack, Gmail, Google Calendar, Drive, Sheets, Microsoft Teams, Salesforce, Shopify, WhatsApp, Stripe, Zendesk and more are coming!
If you need any extra apps that is not in the current sandbox, let us know! We will clone that for you!
What makes it different: • Same API, same behavior: we check each replica against the real app's API, edge cases included (error codes, pagination, rate limits etc). • Every replica has its own UI, so you can open "Slack" or "Gmail" and see exactly what your agent did. • Break things on purpose: inject rate limits and server errors to see how your agent handles a bad day. • Every run starts from a fresh copy, and nothing ever reaches a real customer.
Easy to connect, swap your production URL for your sandbox URL, that’s it!
One question for the community, which app would you never let your agent touch untested?
About StateWeave on Product Hunt
“Test your AI agent on working clones of the apps it uses”
StateWeave launched on Product Hunt on November 16th, 2026 and earned 98 upvotes and 17 comments, placing #12 on the daily leaderboard. StateWeave runs your AI agent in sandboxes of the apps it works in: stateful clones of Slack, Gmail, Salesforce, Shopify, Zendesk, Google Calendar, Sheets, Drive, WhatsApp and Teams, with the same APIs, MCP servers and UIs as the real thing. Describe a scenario to fill every app with consistent data, inject faults like outages and 401s, and get graded runs showing which checks failed and the exact calls behind them. Fresh copy every run.
StateWeave was featured in SaaS (45.4k followers), Developer Tools (523.2k followers) and Artificial Intelligence (483.9k followers) on Product Hunt. Together, these topics include over 295.3k products, making this a competitive space to launch in.
Who hunted StateWeave?
StateWeave was hunted by Aleksandar Blazhev. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how StateWeave stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Excited to hunt StateWeave today.
StateWeave helps teams test AI agents on working copies of the apps they actually use, such as Slack, Gmail, Salesforce, Shopify and Zendesk and more.
Instead of relying on mocked APIs that only return what you wrote, StateWeave lets your agent run against twins with the same APIs, MCP servers, web UIs and errors as the real apps, filled with a realistic scenario, so you can see what it gets wrong before it ships.
What stands out:
- Working copies of real applications, with your agent’s existing client libraries unchanged
- Custom app twins on demand: tell the team which app you need, and they'll build a working replica for you
- Generate a whole consistent company or scenarios (people, history, records) from one sentence, or bring your own data.
- Inject real-world failures like outages, expired tokens and cut-off responses
- Grade every run and compare agent versions side by side
- Let Claude Code, Codex or Cursor run the tests, read what broke and fix it