Ostra overview
Ostra is TestRelic's family of testing models — Ostra Decision for fast CI decisions (failure triage, test impact, selector survival) and Ostra Pro for writing, fixing and analysing tests — available in the TestRelic Console, over REST and over MCP.
Ostra is TestRelic's family of testing models: models built for the questions test engineers and CI pipelines ask every day, priced per use, and reachable from your code, your pipeline and your AI coding agent.
The models
| Ostra Decision | Ostra Pro | |
|---|---|---|
| API id | ostra-lite | ostra-pro |
| Status | Beta | 1.0 |
| Answers | Three CI questions, each with a class and a confidence | Open-ended testing work, in text |
| Speed | Milliseconds | Seconds; streams as it writes |
| Price | Per task, from $0.25 per 1M | Per token |
| Best for | Running on every failure, every diff, every release | Writing and fixing tests, analysing failures, reading screenshots and video |
Ostra Decision
Ostra Decision is one model with three tasks:
| Task | Question it answers | Answers |
|---|---|---|
| Failure triage | Why did this test fail? | flaky, likely_flaky, product_bug, test_bug, environment, selector_drift or defer |
| Test impact | Which tests does this change affect? | impacted or not_impacted per test, plus the list to run |
| Selector survival | Will this locator survive the next release? | survives or breaks per locator |
Every answer comes with its probability, and the model says defer instead of guessing when it isn't confident. It's built to run on every failure in every pipeline: answers come back in milliseconds and cost fractions of a cent per thousand. See Ostra Decision.
Ostra Pro
Ostra Pro is a testing model you use the way you'd use a general-purpose assistant: ask it to write a Playwright test, explain a 2,000-line CI log, review a test diff, turn a bug report into test cases, or compare a screenshot with the design. It reasons before answering (you choose how hard), calls your tools, reads images and video, and works with OpenAI and Anthropic clients unchanged. See Ostra Pro.
Coming soon
| Ostra Code | Production-ready test code from a page's DOM, for any framework. |
| Ostra Visual | Visual regression with high-precision DOM change detection and results. Supports video. |
| Dedicated deployments | Reserved Ostra Decision capacity on a private endpoint. Today every request runs on the shared serverless pool. |
Join the waitlists from Models and Deployments in the TestRelic Console.
Ways to use Ostra
| Where | What you do | Start here |
|---|---|---|
| TestRelic Console | Try both models in the Playground, create API keys, add credit, watch usage and logs. | Console overview |
| REST API | Call the models from any language: https://ostra.testrelic.ai/api/v1. | API reference |
| MCP server | Give Claude Code, Cursor, VS Code or any MCP client the models as tools: https://ostra.testrelic.ai/mcp. | MCP server |
| OpenAI and Anthropic SDKs | Point an existing client at Ostra Pro by changing its base URL. | OpenAI and Anthropic clients |
One API key works for both models, every task, REST and MCP.
Pricing
There are two ways to pay for Ostra. The Growth, Pro and Enterprise plans include Ostra usage, metered and reset automatically; see pricing. Without a plan, or for your own API keys, Ostra is pay as you go and prepaid through the Console: you add credit to a workspace ($5 to $999 per top-up, by card), and each request is charged against it. The Console has no subscription, no free tier and no invoice, and its credit is separate from a plan.
| Ostra Decision | Billed per | Per 1M | With /batch |
|---|---|---|---|
| Failure triage | Failure | $0.50 | $0.25 |
| Test impact | Request | $2.00 | $1.00 |
| Selector survival | Locator | $0.25 | $0.125 |
| Decisions returned | $0.00 | $0.00 |
| Ostra Pro | Per 1M tokens |
|---|---|
| Input | $0.64 |
| Cached input | $0.06 |
| Output (reasoning included) | $4.47 |
- Identical Ostra Decision requests within 24 hours are answered from cache for free.
- Requests that fail cost nothing.
- A monthly spend limit per workspace and a monthly cap per key stop surprises.
Full rules: Billing and usage and Billing and credits in the Console.
Usage levels
Rate limits are per workspace, in requests per minute, and rise automatically with the credit you've bought:
| Level | Total top-ups | Requests per minute |
|---|---|---|
| 1 | First top-up ($5) | 600 |
| 2 | $50 | 2,000 |
| 3 | $250 | 6,000 |
| 4 | $1,000 | 20,000 |
Coupon credit doesn't count toward the level. See Rate limits.
Your own environment
Both Ostra Decision and Ostra Pro can run in air-gapped (on-premises), bring-your-own-cloud and managed-hosting environments, set up by arrangement with the TestRelic team and included in Enterprise. See TestRelic Enterprise or talk to us to plan one.
Next steps
- Quickstart: from sign-in to your first answer in five minutes.
- Ostra Decision and Ostra Pro: what each model does and how to get good answers.
- Guides: put Ostra in CI and in your coding agent.