Pome
Log inSign up

Verified sandboxesfor AI agents.

Your agent runs against digital twins of GitHub, Stripe, Slack, Gmail, and Linear — graded against the live API every day. Every run comes back scored.

pome· Customer Support Agent2440ms
Current run: Agent is processing a $150 refund with open chargeback (double pay issue)
SpanTimeline · 0 → 2440 msDuration
Run: Double Pay Error2440ms
prompt → agent.invoke108ms
llm.plan · triage382ms
gmail.messages.get124ms
stripe.charges.get168ms
fraud.check → low_risk210ms
llm.plan · approve $150398ms
stripe.refunds.create286ms
criterion · dispute8ms
9 of 10 tasks passed
Failed · refund-chargeback-double-pay — agent issued a $150 refund on a charge with an open Stripe chargeback.
Open in Pome ↗
Start locally

Your first run is local, and needs no account

One command boots a digital twin on your own machine. It prints its MCP URL and its own token, so you can point an agent at a real GitHub, Stripe, Slack, Gmail, or Linear surface before you sign up for anything. The twins are Apache-2.0 and the runtime is public.

npx @pome-sh/cli twin start github
twin listening on http://127.0.0.1:3333MCP http://127.0.0.1:3333/mcpPOME_AUTH_TOKEN printed here — no login, no key to fetch

Needs Node.js 24 or newer. Everything ships in @pome-sh/cli. Grading a run against pass/fail criteria is the hosted half.

Bring production-level evals to your development workflow

Start a sandbox and run your agent against digital twins of the real APIs at every stage of development. Test edge cases that track API changes and production failures.

Write tasks

Describe what your agent should do and how it should do it. Pome runs each task against digital twins of the APIs it calls.

TESTS.md
# Agent test tasks
## Test 1
### seed:
### success:
## Test 2
### seed:
### success:

Watch the agentic “flight recorder”

Every tool call and state mutation is logged into a replayable audit trail. Rewind and debug multi-step failures that standard observability misses.

trace · run_4f2fail
agent.start
tool.lookup
llm.plan
tool.write
commit

Disable destructive actions before production

Surface every destructive action from production traces. Toggle off unauthorized calls. Past tasks inform future runs to prevent regressions.

GitHub Agent
github.com
3 allowed2 denied
reversible
GETlistIssues
POSTaddComment
irreversible
DELETEdeleteRepo
POSTpulls.merge
AI agent
Pome
GitHub
5 digital twins · 115 MCP tools · every route graded daily
githubLive
stripeLive
slackLive
gmailLive
linearLive

Verify everything

Every run is graded on what actually changed inside the digital twins — exact assertions on tool calls and state, not an opinion about the transcript.

01

Code graders

DeterministicAssertable

Assert on twin state and tool calls after each run. Identical seeds produce identical results, so you can gate merges in CI. This is what grades a run by default.

02

Model-judged scoring

BetaOpt-inNon-deterministic

For open-ended behavior you can opt a task into a judge model that scores the run against your rubric and cites what failed. It is in beta, it is off unless you ask for it, and it is not what gates a merge.

Ready to try it

Simulate agent failures before your users see them

Book a demo for a walkthrough or try it yourself.

Frameworks & agents we support

Claude Managed AgentsLangChain EcosystemVercel AI SDKOpenAI Agent SDKClaude CodeCursor

Build reliably with Pome

Digital twins and replayable traces — so every agent rollout ships with proof, not best effort. The meter counts sandboxes, the thing you start.

Free
$0
  • 3 concurrent sandboxes
  • 100 agent evals / month
  • 5,000 MCP / API calls / month
  • 1 trial per task
Sign up
TeamMost popular
$500/ month
  • 25 concurrent sandboxes
  • 10,000 agent evals / month
  • 1 Million MCP / API calls / month
  • Up to 10 trials per task
  • Onboarding + priority support
Book a demo
Enterprise
Custom
  • Everything in Team
  • Self-host option
  • Custom digital twins
  • Custom evaluation framework
  • White-glove support
Talk to sales

Need SSO, self-host, or a custom contract? Contact us →