Define agentic check runs in markdown. Macroscope's research agent enforces them on every pull request — with full access to your codebase, git history, and connected integrations.
Start with $100 in free usage and 1,000 free agent credits every month.
Define your team's rules in markdown. Macroscope enforces them on every PR — automatically.
$100 in free usage and 1,000 free agent credits every month. No seat fees.
A check run agent is a custom, agentic check run that enforces your team's rules on every pull request. You define it as a markdown file in .macroscope/check-run-agents/ with natural-language instructions. The agent runs automatically on every PR — with access to your full codebase, git history, and connected integrations like Slack, Sentry, and Jira.
Macroscope's built-in Correctness check run reviews every PR for bugs and issues automatically. Check run agents are additional, custom checks you define — they can enforce team-specific standards, check for security patterns, verify API contracts, validate event tracking, or anything else you can describe in plain English. They run alongside the built-in reviews.
By default, agents can browse your code, run git commands (log, blame, diff, grep), read GitHub metadata (issues, labels, PR info), and modify the PR (post comments, add labels, assign reviewers). You can also grant them access to Sentry, PostHog, Slack, Jira, Linear, BigQuery, LaunchDarkly, Amplitude, GCP Cloud Logging, and any MCP-compatible server.
Yes. Set conclusion: failure in the agent's frontmatter, and it becomes a blocking check. When the agent finds issues, the GitHub check reports a failure status, preventing the PR from being merged. For advisory-only feedback, use conclusion: neutral (the default).
Check run agents support Claude Opus 4.5, 4.6, and 4.7, Claude Sonnet 4.5 and 4.6, and GPT-5-2, 5-4, and 5-5. You choose the model per agent. For Anthropic models, you can also tune the reasoning depth (off, low, medium, high) and effort level (low, medium, high). Claude Opus 4.6 is the default.
Check run agents are billed as agent credits at $0.01 per credit. Every workspace gets 1,000 free credits per month (~$10 value). New workspaces also receive $100 in free usage to get started. You can control costs by scoping agents with include/exclude patterns, choosing the right effort level, and using full_diff mode instead of the more expensive code_object mode.
Create a .macroscope/check-run-agents/ directory in your repository, add a markdown file with your check definition, and commit it to your default branch. The agent will start running on the next PR. It takes under 5 minutes to set up your first check run agent — no configuration beyond the markdown file is needed.
Join teams building with Macroscope to catch more bugs and merge PRs faster than ever. Simple usage-based pricing. No seat fees, no surprises.
