Build your own software factory

Run fleets of coding agents in parallel, automatically catch bugs and validate fixes, and see what’s happening across your teams.

Cars24
Parallel
UnitedMasters
Lightfield
Tilt
Pydantic
Particle
Imprint
Casa
Customer logo
beta

Murmur

Cloud Agent Orchestration

Run dozens of agents in parallel without babysitting them. Every agent writes code and verifies its own work in an isolated sandbox.

Self-driving PRs

Murmur listens to GitHub and automatically deals with code review comments, failing checks, and merge conflicts.

add-system-theme-option
PR Opened · #13868
PR Comment · Changes requested
CodexWorking
Reading Macroscope review
Send a message…

Sandboxed

Each agent runs in a dedicated cloud VM that can run your full stack. This allows agents to self verify their work, and to run longer since they're untethered from your local machine.

turn-follow-up-composer8 vCPU · 16 GB · eu-westrunning
workspace-agent-presence2 vCPU · 4 GB · us-westbuilding
desktop-shell-polish4 vCPU · 8 GB · us-eastrunning
Repositories
FrontendBackend
Running stack3 services

web

:3000

api

:8080

postgres

:5432

$ pnpm test$ playwright screenshot$ git push
Live preview
preview:3000live
Home
Search
IS

Executive summary

A clear view of what changed across your workspace.

Jun 3–16

Sprint Summary

Session middleware refactorConsolidated auth checks into a single middleware layer and removed deprecated token handlers from the request path.
Ledger reconciliationMatched payment events against ledger entries and tightened retry handling for disputed or delayed transactions.
Observability stackExpanded service traces and dashboards around CI, secrets rotation, and queue reliability work.
Onboarding funnelImproved the first-run workspace path and referral surfaces for new growth experiments.

Local orchestration

Use your local machine to direct a fleet of cloud agents on Murmur using the CLI/MCP.

> fan out murmur agents to finish the homepage redesign
I'll split the work across isolated sandboxes. ⎿ Ran murmur spawn 6 --repo back ⎿ 6 sandboxes ready (claude ×4, codex ×2) ⎿ murmur status 4 running, 2 queued ⎿ homepage-hero opened PR #436 checks passing
✻ Formicating… (esc to interrupt)

Bring Your Own Keys

Use your Claude or Codex subscription and manage your own API keys.

Coding Agent API Key

encrypted
Choose an agent provider
sk-ant-••••••••••••7W2Eencrypted
Scoped to agent execution in this workspace

Run on our cloud or yours

Choose the execution placement for every workload. Host VMs in Murmur Cloud, or manage them in your own GCP or AWS VPCs.

Runtime placement

ready
Choose where Murmur runs agents
Murmur managed computeDedicated cloud workers, managed for you.
No infrastructure to manage.

Works where you work

Murmur integrates with Slack, Linear, GitHub, or your own systems, making it easy for teams to coordinate agent work.

Security & customization

Choose each VM's CPU, memory, and disk; bake your full stack into a custom image; and tune warm capacity and termination behavior.

Production workspaceImage recipe
Base imagemurmur-debian12
Recipeprovision.sh
Baked imagemacroscope-devcustom runtime artifact
CPU4 vCPU
Memory16 GB
Disk120 GB SSD
Min warm VMs2
TerminationImmediate

We currently only support GitHub source control. We also recommend investing time to make your development environment sandbox-ready (so agents can use/modify/break services without disrupting your staging and production environments) to maximize your mileage with Murmur.

Automatic review and validation

Code Review

Catch real bugs

Macroscope finds more bugs than any other code review tool, without spamming you with false positives.

Approvability

Take PR approvals off of your team's plate. Macroscope can auto-approve PRs that have no bugs and have minimal blast radius.

Clarify warning copy in checkout banner#1842

Correctness Checks

Out of the box, Macroscope finds more bugs than any other tool, without spamming you with false positives

05session: Session
06) {
07if (session.expiresAt <
08Date.now()) {
09return true
10}
11return false
12}
13
14export async function createSession(
15userId: string
16) {
17const token = generateToken()
18const expiresAt = Date.now()
19+ (60 * 60 * 24)
16export async function createSession(
05session: Session
06) {
07if (session.expiresAt <
08Math.floor(Date.now() / 1000)) {
09return true
10}
11return false
12}
13
14export async function createSession(
15userId: string
16) {
17const token = generateToken()
18const expiresAt = Math.floor(
19Date.now() / 1000) + 86400
16export async function createSession(
2 bugs detected by Macroscope
auth/session.ts

Critical Bug · Line 07–08

Date.now() returns milliseconds but expiresAt is stored in seconds. Every session will appear expired immediately after...

What makes our code review different?

We combine an agentic pipeline with purpose-built Abstract Syntax Tree (AST) walkers that we’ve built for the most common programming languages. This allows our agent to deterministically gather rich context about every change, without relying solely on an LLM to search for that context itself. The result is faster, more token-efficient analysis and a deeper understanding of the codebase. This is one reason we achieve higher precision at 1/20th the cost of Claude’s Code Reviewer, according to independent benchmarks. On Martian’s benchmark, read 9 September 2026, 79.5% of Macroscope’s suggestions matched a change the developer actually made, against 69.4% for Claude.

Source: Martian benchmark and official Claude pricing post.

48.31% Recall
~$1/review
40.9% Recall
~$15-$25/review

Macroscope CLI

Review code locally and catch real bugs before you push.

By the Benchmarks

See how Macroscope compares to other tools according to our internal benchmark.

Read blog

Check Run Agents

Define custom AI reviewers that automatically run within your PRs (each with access to all your connected tools or MCPs), enforcing your stylistic conventions and workflows at PR review time.

Check run agents in progress...
5 agents · defined in .macroscope/check-run-agents
Security Review
Web Event Tracking
Ticket Requirements
Accessibility Audit
Production Errors

PR Summaries

Every PR gets a clear summary with context for reviewers.

Fix Shopify route page title fallbacks

PR #1738 · OpenJordan wants to merge intomain
Summary

This PR fixes checkout titles for Shopify routes whose projected slug is flattened before the title resolver runs.

  • Adds resolveRouteTitleFallback to recover store titles when a route exists but the generic resolver returns an empty value.
  • Keeps non-Shopify routes on the existing title resolution path.
  • Adds coverage for flattened slugs, nested product paths, and missing store metadata.

Start review in src/routes/titleResolver.ts, where the fallback is applied after projection.

Written by Macroscope

Fix It For Me

Reply "fix it" on any review comment. Macroscope branches, writes the fix, runs your CI, and merges when checks pass.

Jordan Mills

Removes the settlementDate guard from invoice replay. Paused accounts can now be charged twice when the retry job replays a partial billing window.

Macroscope

This drops the settlementDate guard before invoice replay. Paused accounts can be charged twice during partial-window retries.

@Macroscope Fix it

Scalable Config

Per-repo enable/disable, skip-by-label, file exclusion via .macroscope/ignore, manual triggers via @macroscope-app review. Personal preferences override workspace defaults.

CLI

beta
curl -sSL https://raw.githubusercontent.com/prassoai/macroscope-local/main/install.sh | bash

Run Macroscope's correctness review locally on your machine and catch real bugs before you push to your branch.

macroscope — zsh

/macroscope:code-review

Reviewing staged changes against main…

Correctness4 issues · auth/session.ts

I've fixed all four issues in a parallel worktree — want me to apply them?

✻ Cooked for 4m 18s

works in your agent

CodexCodex
Claude CodeClaude Code
CursorCursor
OpenCodeOpenCode
Automatic visibility

Status

Know what's shipping. Without asking.

Engineering Productivity

Understand engineering output across your team. See how much engineering work is actually landing in production.

Pushed vs. landed

Team Members
All
Project
All
Date Range
May 22-29

Integrations

Native — Deep, first-party integrations that power Code Review, Status, and Agent.

Macroscope

Macros — agentic automation

Trigger recurring workflows, with guardrails, from inside your stack.

API — wire it into your stack

Connect Macroscope to your CI, chat, support, and observability tools.

A newsfeed for code changes

Exceptionally succinct and accurate summaries of each commit and PR, automatically distributed to your team in Slack.

Alex Riverajust now

Added webhook support for payment retries and failed deliveries.

3h 30mcoding time
Angie Nunez20m ago

Implemented streaming chatbot responses with inline fallback handling.

2h 10mcoding time
Jordan Mills46m ago

Integrated Stripe usage billing with invoice finalization webhooks.

5h 40mcoding time
Helen Winstonship1h 12m ago

Moved notifications to a queue-based delivery system.

3h 50mcoding time
Marco Devlin2h 05m ago

Added cursor pagination and indexes for audit logs.

1h 25mcoding time

Executive Summaries

High-level summaries of what's changing in your codebase, regenerated multiple times daily and surfaced on the home page and via email. Organized by Area, collapsible, scannable in seconds.

Sprint Summary

Session middleware refactorConsolidated auth checks into a single middleware layer and removed deprecated token handlers from the request path.
Ledger reconciliationMatched payment events against ledger entries and tightened retry handling for disputed or delayed transactions.
Observability stackExpanded service traces and dashboards around CI, secrets rotation, and queue reliability work.
Onboarding funnelImproved the first-run workspace path and referral surfaces for new growth experiments.

Slack — answers and actions

Use it directly from Slack. Inspect code, diagnose failures, and ship.

See it in action

How our team uses Macroscope on real work.

Your software factory, in motion.

Orchestrate coding agents in parallel, review every change automatically, and keep work moving from task to merge.

Start For Free

Security

Keeping your data secure is of paramount importance to us. We have rigorous technical, operational and contractual safeguards in place to protect customer data.

SOC 2 Type II compliant

Visit our Trust Center to request copies of our audited documentation.

Visit Trust Center

Encrypted in transit and at rest

All customer data is encrypted at rest and in transit.

Architecturally isolated

Customer Code is architecturally isolated and secured by design.

Zero model training

Macroscope does not train models using your source code, and our agreements with model providers ensure they do not train using your IP.

People seem to like it...

Parag Agrawal
CEO, Parallel

We are huge fans. It's the best code review tool we've tested and I personally love that we can avoid unnecessary status update meetings.

Scott Belsky
Cofounder of Behance @ Founder A24 Labs

Macroscope has become a core part of our engineering team, bringing some new superpowers in productivity and keeping us all aligned and up to speed on what's getting done every day.

Kaz Nejatian
CEO, Opendoor

I filed a minor bug with @Macroscope at 9 am London time. By 3 pm London time (7 am SF time!) @kayvz personally responded and was fixing it. Founder mode.

Logan Fisher
CPO & CTO, Parkhub

Macroscope just gets it. They've cracked the code on turning huge amounts of complex data into clear, actionable insights-and they've made it effortless for us to make smarter decisions, faster.

Shane Mac
CEO, XMTP Labs

Rarely do I see a product that helps everyone in the org save time and do more without asking us to do more. The team does less updating while leadership can see what's actually happening - and even ask questions about it - all without distracting anyone.

Michael York
CEO, Casa

Macroscope's commit summaries – drawing from their deep understanding of our projects and codebase – are the beating heart of our engineering organization. The kind of upgrade you can never imagine working without.

Nick Molnar
CTO, XMTP Labs

We've used just about every AI-driven PR assistant out there: the signal to noise from Macroscope is the best I've seen. The PR descriptions are better than what we would have written by hand, and when it flags an issue it's almost always a real bug.

Hardik Patil
Product Lead at Anon

Macroscope makes it easy for me to stay on top of engineering progress and developer productivity, even as a solo product lead working with 8 engineers. Their support is ridiculously responsive, and they're super fast at shipping new features I requested.

Marcel Molina
CTO & Cofounder, Particle

Macroscope is like having a distinguished engineer tech lead who's read every diff, understands every project, and can answer any question about your codebase instantly. We can finally focus on shipping instead of process.