introducing Policy Repo: write your code review config once, and every repo inherits it. pick a Policy Repo under settings → repos → policy inheritance. macroscope then uses the ".macroscope/" config from that repo's default branch as the shared default for all your repos: ✅ "approvability​.md": your policy for which PRs macroscope can safely approve without a human 👻 "ignore​.md": the files code review should skip, so reviews stay fast and focused 🤖 "check-run-agents/": agentic PR checks you write in plain markdown 🎯 "correctness/": tune how macroscope's correctness check works with your own rules and filters need something different in one repo? add that file to the repo's own ".macroscope/" folder and it takes priority over the inherited one. that means no more copy-pasting config into every repo, no drift when one copy gets updated and the rest don't, and one PR to change review policy everywhere. docs: docs.macroscope.com/code-rev…

introducing the Approvability Dashboard: see what macroscope's approvability agent is approving, what it isn't, and how your policy changes move the numbers. approvability works out of the box with sensible defaults. want it more permissive or tuned to your team's standards? add your own rules in "approvability​.md", then watch how those changes play out on your real PRs: ✅ approval rate: the share of PRs macroscope approved, plus a ceiling for what it would have approved 🛡️ eligibility: how often the agent judged a change safe to approve without a human 🚦 verdict mix: approved, would approve, needs human approval, skipped 📋 every PR and its verdict, searchable by repo and time range, and exportable to CSV find it under code review → approvability in the macroscope app.

3/ check run agents are agentic PR checks you write in plain markdown. they now follow the same ignore rules as code review, so they skip test files, .json, or anything else you've told Macroscope to ignore. that means faster checks, simpler configs and much more cost efficiency. if one of your agents does need to review something you've deliberately ignored, use the "include" param in its config.

2/ new: onlyApprove. list paths in approvability's front matter, and it will only consider approving PRs whose diff stays inside them. touch anything outside those paths and it won't approve. so you can turn on auto-approval for the parts of your codebase you trust it with, without handing it the keys to everything.

1/ a few upgrades to approvability and check run agents this week. approvability is a macroscope agent that decides whether a PR is safe to merge without a human, and approves it if so. it helps make code review less of a bottleneck for your eng team. as of this week, when you resolve the last blocking comment on a PR, approvability automatically reruns. so a fixed PR gets approved right away, instead of sitting there waiting on a new commit or a manual rerun. this means faster approvals and faster merges.

Back for day 2 of ELC! Loved all the conversations yesterday and are excited for more today. If you’re at the conference, come say hi :)

Yesterday's #SFTechWeek Clickety Clack Café, in photos 📷 Thanks for stopping by! And if your PR needs another shot, check out Macroscope: macroscope.com

PSA: free coffee and matcha at 500 Folsom St until 4:30pm 🍵 Grab a hot (or cold) fix before your next #SFTechWeek event. We'll be here (caffeinated).

Clickety Clack Café is live!☕️ Coffee & code review on Macroscope. Register for location: partiful.com/e/1XFyBiTbjMeIm… #SFTechWeek

Tomorrow: free coffee for everyone at #SFTechWeek. Clickety Clack Café, 1:30 to 4:30pm. RSVP for the location. Coffee & Code review on us. partiful.com/e/1XFyBiTbjMeIm…

🥉 GPT 6.1 Sol debuts at #3 on our code review leaderboard. It comes below GPT 6 Sol because recall is lower, but it leaves 30% fewer comments with very high precision. It also slightly beats workhorse Astra at a seventh of the price! See full results: macroscope.com/benchmark

@cars24india is one of India's largest auto-tech companies: 300M+ users, $1B+ in revenue, 400 engineers. with Macroscope, they saw up to 3x more code reaching production in 3 months. 🧵

We're opening a coffee cart for #SFTechWeek ☕ Free coffee and matcha for everyone shipping code Tue Oct 6, 1:30 to 4:30pm, downtown SF. RSVP for location drop. partiful.com/e/1XFyBiTbjMeIm…

Building on the release of MacroscopeBench, we’ve been researching post-training our own models at Macroscope, leveraging training infrastructure from @FireworksAI_HQ . Our co-founder, @Rob_Bishop, and Vivek Chauhan, product lead for training at Fireworks, will be sharing learnings from the collaboration on Sep 29, 11am PT.

There are some new top models leading our code review benchmark. 🥇 Claude Opus 5.5 — 82.3 (f1 score) 🥈 GPT-6 Sol — 80.3 🥉 GPT-6 Astra — 78.0 See full results here: macroscope.com/benchmark

Learn more about how we built MacroscopeBench and how it works here: macroscope.com/blog/macrosco…

we rewind to the commit that introduced each bug and ask whether a reviewer would have caught it there. it's never told a bug exists. see which models currently perform the best at macroscope.com/benchmark

Production is the worst place to find out if your code reviewer is good enough MacroscopeBench measures it first, against real bugs that shipped, just not in your repo. It is now the only trusted benchmark for code review on @FireworksAI_HQ SII Index fireworks.ai/specialized-int…

Will you run your Murmur agents on an iPhone Duo?

ad astra. GPT-6 Astra is now available in Macroscope Check Run Agents!

Fable 5.1’s comments skew toward the bugs that matter most. It finds 6.72 critical+high bugs per 100 comments vs. 3.39 for Sol: nearly 2x the signal. The vast majority of Sol’s recall edge comes from extra low-severity issues Fable declines to report.

Happy Fable 5.1 day 🎉 Now available in Check Run Agents. We benchmarked Fable 5.1 alongside other frontier models, and we’re very impressed. The most notable thing about the results below is that with this benchmark, our harness and prompts were still tuned for GPT-5.6 models. And yet, Fable 5.1 inched just behind GPT‑5.6 Sol (high) for the highest recall of any model we’ve ever measured—and it does so by posting half as many comments and gets there with roughly 60% fewer tool calls. We’re in the process of running both GPT-5.6 Sol and Fable 5.1 on a neutral prompt and will follow up with updated results when we have them.

Macroscope Check Run Agents now support GLM-5.3! This model supports large-context reviews and configurable reasoning levels, so teams can tune review depth and cost for their workflow.

New models just dropped in Macroscope Check Run Agents: Grok 4.6 and Qwen3.8-Max. Try them out and let us know what you think!

Each Check Run Agent is a Markdown file in your repository that defines a custom AI agent. You can choose the model that fits the scope and priorities of each task, and write instructions just as you would for an agent skill. Once the file is merged into your default branch, Macroscope runs the agent as a GitHub check run on every PR that matches its file scope.

Macroscope Check Run Agents now support DeepSeek-V4-Flash! We’ve seen major improvements to the model’s agentic capabilities. In our internal evaluation, DeepSeek-V4-Flash delivered near-frontier bug detection at a much lower cost per task, though on average it took 2–5× longer to complete a review compared with GPT-5.6 sol and Opus 5.

Check Run Agents now support Kimi K3. With performance rivaling top frontier closed models, K3 can power your custom code review checks at a fraction of the price of Fable 5 or GPT-5.6 Sol.

Claude Opus 5 is now available as a Check Run Agent in Macroscope!

Until now, getting a Macroscope code review required opening a PR. Today, that changes with the Macroscope CLI. Get fast, high-signal bug detection directly from your terminal—before you push your code. Just install the CLI, then: - Ask your agent to trigger a review with /macroscope - Run “macroscope codereview” to trigger a review in the TUI

Congratulations to @jonathanfishner for taking the top spot in our challenge and taking home the drone🥇

We've got a challenge at our AGI Summit booth today 60 seconds. Bug or no bug. Highest score by the end of the conference wins a Drone!

Find our bug exterminators for some freshly caught bug candies!

We're at the AGI Summit today at the Palace of Fine Arts in SF. Swing by our booth, say hi, and maybe even win a drone!

Check Run Agents now support these open source models: • kimi-k2-7-code • glm-5-2 • minimax-m3 • qwen-3-7-plus The open source model ecosystem has improved dramatically, and many OSS models are now competitive with frontier models across coding and reasoning tasks. We built Check Run Agents around a simple idea: expose as much power and flexibility as possible while keeping the experience dead simple. Broad model choice is a natural extension of that. See our docs for the full list of supported models, pricing, and configuration instructions.

fable 5 and sonnet 5 are now available as check run agents!

Macroscope Agent reacts to events across your workflow and takes action autonomously. In this demo, Agent: → Pulls PostHog data to answer product questions in seconds → Auto-updates docs and enforces feature flag rules when PRs open → Catches a Sentry error, traces the root cause, and opens a fix PR with tests

Agent now has an API. Send a POST with a query and where you want the answer. Agent runs async and delivers to Slack channel, a thread, or a custom webhook. Connecting any workflow: CI pipelines, incident tooling, internal apps. If it can make an HTTP request, it can use Agent.

Agent can run on demand, on a schedule, and now based on event triggers. Macros can trigger on new commits and new PRs. They get full context about what changed and can filtered to only run on certain repos or branches. Deliver results to Slack channels, Slack threads, or webhook URLs.

With MCP connections, you can connect it to third-party services that we don’t already have native connectors for. If it speaks MCP, Agent can use it. Here are some that we use:

Agent now connects to more of your stack: → Sentry — search issues, pull event details, list affected projects → Amplitude — event segmentation, funnels, retention, user activity → GitHub API — PR metadata, issue history, workflow runs, review comments, repo stats → GCP Cloud Logging — production logs by service, severity, time range → Any MCP server — header or OAuth auth, per-tool access controls ...in addition to all existing capabilities: code research, git tools, code gen, PR creation, Slack, Linear, Jira, BigQuery, LaunchDarkly, PostHog, and more.

Ask Macroscope Anything (AMA) is now Macroscope Agent, and it got a big upgrade today: → 6 new integrations (Sentry, Amplitude, GCP Logging, GitHub API, image gen, and MCP) → Event-driven automations triggered by commits and PRs → Webhook delivery + an API

How it works: You’ll prepay via credits, which are then consumed as usage occurs. Credits are auto-refilled according to your Refill settings. You can also control your spending with flexible settings: → Set spend limits per review, per PR, and per month → Disable reviews on specific repos, PR authors, or file types → Only allow reviews when manually invoked

Loved seeing all the projects using Macroscope’s new features at the Deep Agent hack today! 💻

On @withmartian’s independent Code Review benchmark, Macroscope ranks #5 overall on f1 score (a metric that balances precision and recall) and higher than formidable competitors we respect like Cursor BugBot, Devin, Greptile, CodeRabbit, and Claude Code Reviewer

Deep codebase understanding means an agent that (without being told): > knows your feature branch exists > knows the bug lives there > targets the fix correctly Fix bugs from anywhere with Macroscope

You can invoke Macroscope from anywhere using webhooks (@zapier , @n8n_io , your own APIs). All you need is a prompt + destination > Macroscope will post the result there.

Macroscope's review comments in @github include a copy/paste-able prompt for your favorite coding agent. See the issue > paste the prompt > fix it.

You can now create PRs directly from Slack with Macroscope. Customer reported a bug via screenshot? Just @ tag Macroscope → 👀 it reads the Slack thread → 📄 parses any attachments → ✅ creates a PR fixing the bug

GitHub review comments now include a copypaste-able prompt for your coding agent. See the issue. Paste the prompt. Fix it.

We've also added direct integrations with BigQuery, @LaunchDarkly, & @PostHog. Macroscope can now extract context from your feature flags, web analytics, and datasets.