# Mirrors

> Mirrors is the staging environment your AI agents never had. It rebuilds the systems they call, then replays real sessions against each prompt, tool, or model change before it reaches production.

Mirrors is a developer platform for testing and evaluating AI agents. It turns traces,
agent code, tool code, or docs into a runnable staging environment: a seeded database
plus bound tools that behave like the systems the agent calls. Agents replay against
that environment, never live systems, so destructive flows such as refunds, deletes,
and sends are safe to test.

## How it works

1. **Provide what you already have** - traces, agent code, tool code, or docs. The optional mirrorkit collector keeps the environment current from live sessions.
2. **Mirrors builds the environment** - schema, fabricated seed data, and tool behavior become a runnable copy of the systems the agent calls.
3. **Replay any session against any change** - use the dashboard, CLI, MCP server, public API, or CI to compare versions without touching production.

## What you can do with it

- Catch regressions before they ship: replay real sessions against every change and see exactly what would have broken.
- Reproduce the bug that paged you on demand in a stable, seeded environment.
- Test risky and destructive flows safely against the environment instead of live systems.
- Gate CI on replays so every change is checked before it merges.
- Measure whether an agent change is actually better with per-tool coverage and accuracy scores.
- Drive mirrors from your own code over a versioned /v1 API authed by workspace keys.

## Pricing

- **Free** - $0/month: 60 replay-minutes each month, then $0.20 per replay-minute; unlimited environments, CI gates, and the public /v1 API.
- **Startup** - $249/month: 2,000 replay-minutes included, then $0.15 per replay-minute; priority support and onboarding.
- **Enterprise** - custom volume pricing with SSO, role controls, security review, and on-premises or BYOC options.

## Key facts

- Category: AI agent staging environments, testing, and evaluation.
- No production access is required during replay; agents run against the environment only.
- The environment keeps the shape of entities and relationships while fabricating values, so redacted sessions work.
- Replay diffing compares each recorded production call with its replayed counterpart to show agreement and drift.

## Data handling

- Customer content (traces, uploaded files, built environments) is never used to train models and is never sold.
- Model calls made while an agent runs are routed with zero data retention requested.
- Content is kept until the customer deletes it; deleting an environment removes its traces, uploads, artifacts, image, and logs. Deletion requests are completed within 30 days.
- Full policy: https://www.runmirrors.com/privacy

## Links

- Home: https://www.runmirrors.com/
- Pricing: https://www.runmirrors.com/pricing
- Contact: https://www.runmirrors.com/contact
- Docs: https://www.runmirrors.com/docs
- Docs index for agents (every page as raw markdown): https://www.runmirrors.com/docs/llms.txt
- Privacy & cookies: https://www.runmirrors.com/privacy
- Get started: https://www.runmirrors.com/login