Augment Code
by Augment Inc.
Agent orchestration for the whole software delivery loop, not a single IDE assistant
Augment Code is an AI software-engineering platform for teams, built around Cosmos, an agent orchestration layer that runs code review, ticket-to-PR, vulnerability remediation and incident triage as continuous loops with human quality gates. It is aimed at engineering organisations rather than individual developers, and is sold as a flat-rate pooled-usage subscription instead of per seat.
Augment Code is built by Augment Inc., founded in 2022 in Palo Alto by Igor Ostrovsky and Guy Gur-Ari, which emerged from stealth in April 2024 with a $227 million Series B at a $977 million post-money valuation led by Sutter Hill Ventures alongside Index Ventures, Innovation Endeavors, Lightspeed Venture Partners and Meritech Capital. The product began as an IDE assistant — a VS Code and JetBrains extension that accumulated more than 770,000 installs — differentiated by a proprietary Context Engine that indexes large codebases so suggestions reflect real project structure rather than the open file. In 2026 the company repositioned around Cosmos, an agent orchestration platform intended to run at organisational rather than individual scale. Cosmos models work as interconnected loops: a code review loop that gives every pull request a substantive first-pass review within minutes, a ticket-to-PR loop the vendor claims delivers two-to-three times the throughput, a vulnerability remediation loop it claims fixes over 60 percent of CVEs automatically, and an incident response loop it claims triages and resolves over 70 percent of alerts before an on-call engineer arrives. Those figures are vendor-reported and unaudited. Pre-built expert templates cover Work Dispatcher, PR Author, Code Review Fleet and Tester roles, shared memory carries learning across runs, and human quality gates pause work for review rather than merging autonomously. Execution is flexible — laptops, developer VMs, managed cloud or customer infrastructure — with CLI access, MCP and native tools, and integrations into GitHub, Jira, Slack and CI/CD. The strategic consequence matters to buyers: the legacy IDE extension is maintained only for existing enterprise contracts, and new customers must adopt Cosmos. Named users include Adobe, MongoDB, Pure Storage, DXC, Snyk, Webflow, Crypto.com and MoneyGram.
A VP of Engineering or platform lead at a mid-size or larger engineering organisation that wants agents wired into GitHub, Jira and CI/CD to absorb review, remediation and triage work — not individual developers looking for a better autocomplete.
Repetitive middle-of-the-SDLC work — first-pass PR review, CVE fixes, ticket-to-PR, alert triage — runs continuously with humans kept at explicit quality gates rather than in the loop for every keystroke.
At a Glance
- Category
- Developer Tools
- Pricing
- Subscription, Usage-based, Flat rate, Contact for pricing
- Target Market
- CTOs, VPs of Engineering, Enterprise Developers, Platform Engineers, DevOps Engineers
- Deployment
- Cloud-first, Hybrid, Self-hosted
- Founded
- 2022
- Headquarters
- Palo Alto, United States
- Team Size
- 51-200
Key Features
- ✓Cosmos agent orchestration
Coordinates multiple agents across triage, authoring, review and verification so work moves through the delivery pipeline instead of stopping at a single chat response.
- ✓Context Engine
Indexes the full codebase to give agents real project structure, which is the capability practitioners most consistently credit for its multi-file refactor quality.
- ✓Pre-built expert templates
Ships ready-made agent roles — Work Dispatcher, PR Author, Code Review Fleet and Tester — so teams start from working configurations rather than authoring agents from scratch.
- ✓Human quality gates
Work pauses at defined checkpoints for human review before proceeding, which keeps engineers accountable for merges while agents do the intermediate labour.
- ✓Flexible execution targets
The same agents run on developer laptops, dev VMs, Augment's managed cloud or entirely inside customer infrastructure, which matters for regulated code estates.
- ✓Shared memory across runs
Learning from previous executions is retained and reused, so repeated tasks in the same repository improve rather than restarting cold each time.
- ✓CLI, MCP and native tool access
Agents reach the toolchain through a command line, Model Context Protocol servers and built-in integrations with GitHub, Jira, Slack and CI/CD pipelines.
Capabilities
Use Cases
- •Automated first-pass pull request review
Every PR receives a substantive review within minutes so human reviewers spend their attention on design rather than mechanical defects.
- •Ticket to reviewed pull request
A backlog ticket is picked up, implemented and opened as a PR with tests, which the vendor claims lifts throughput two to three times.
- •Vulnerability remediation at scale
CVEs across many repositories are patched and verified automatically, with the vendor claiming over 60 percent are remediated without human authoring.
- •Incident triage before escalation
Alerts are investigated and contextualised ahead of the on-call engineer, with the vendor claiming over 70 percent are resolved before escalation.
- •Multi-file refactors in large codebases
The Context Engine identifies affected files and integrates components across a refactor rather than requiring explicit per-file instructions.
Ideal For
Best For
- ✓Engineering organisations that want every pull request to get a substantive automated first-pass review within minutes
- ✓Teams drowning in CVE and dependency remediation work that want fixes drafted and verified automatically
- ✓Converting backlog tickets into reviewed pull requests without a developer starting from a blank editor
- ✓On-call teams that want alerts triaged and context assembled before a human engineer picks up the page
- ✓Large legacy codebases where context quality — not raw model capability — is the limiting factor on AI usefulness
Not Ideal For
- ✗Solo developers and hobbyists — entry is a $100/month flat Business plan aimed at teams, and there is no transparent free trial to evaluate with
- ✗Teams that specifically want an in-editor assistant: the VS Code and JetBrains extension is maintained only for existing enterprise contracts, and new customers are directed to the Cosmos platform instead
- ✗Organisations that need predictable per-seat budgeting, since usage is pooled credits against LLM inference and compute time rather than a fixed seat cost
- ✗Buyers evaluating on published independent reviews — third-party review volume is very low for a company of this funding level, so most evidence is vendor-supplied
Integrations
Deployment
Market Analysis
Pros
- ✓Context quality is the standout: developers report it maintains context across complex refactors and understands project structure where competitors need per-component instructions
- ✓Covers the middle of the SDLC — review, remediation, ticket-to-PR, incident triage — rather than only code authoring, so the buying case is organisational throughput
- ✓Flat-rate pooled pricing at $100/month for up to 50 seats is cheap per developer compared with per-seat coding assistants, if usage stays inside the allowance
- ✓Enterprise controls are strong for this category: SOC 2 Type II, SSO/OIDC/SCIM, customer-managed encryption keys, ISO 42001 and an explicit no-AI-training commitment
- ✓Well capitalised at $252M total funding with named enterprise users including Adobe, MongoDB, Pure Storage, DXC and Snyk
Cons
- ✗The IDE extension that built the user base — 770,000+ installs across VS Code and JetBrains — is now maintained only for existing enterprise contracts, and new customers must move to the Cosmos platform
- ✗Pricing has churned repeatedly: free daily-request allowances were withdrawn, and Augment publicly stated that 22.5 percent of users were consuming twenty times what they paid, which is the backdrop to repeated plan changes
- ✗Pooled usage credits make spend hard to forecast — the $100/month flat fee is a floor that top-ups extend, and heavy agent activity is exactly what the product encourages
- ✗Not viable for solo developers or evaluation-minded teams: the entry plan is team-priced and there is no transparent free trial
- ✗The headline productivity numbers (2-3x throughput, 60%+ CVE remediation, 70%+ incident resolution) are vendor-reported and unaudited, with no independent benchmark available
- ✗Third-party review coverage is thin relative to the company's funding, so buyers have little independent evidence on support quality or long-run reliability
Pricing
Business
From $100/mo (flat)
- ✓$100/mo of pooled usage across LLM, Context Engine and compute
- ✓Up to 50 seats
- ✓Cosmos, CLI access, MCP and native tools
- ✓50 concurrent sessions
- ✓Usage analytics
- ✓Daemon mode
- ✓SOC 2 Type II
- ✓No AI training on your code
- ✓Pay-as-you-go top-ups
Enterprise
Contact for pricing
- ✓Custom user pricing and bespoke usage limits
- ✓Unlimited users and concurrent sessions
- ✓Custom-size and multi-region compute
- ✓SSO, OIDC and SCIM
- ✓CMEK and ISO 42001
- ✓Volume-based annual discounts
- ✓Dedicated support
- ✓No AI training on your code
Pricing is flat-rate and pooled rather than per seat: the Business plan is $100 per month for up to 50 seats and includes $100 of monthly usage across LLM inference, the Context Engine and Cosmos compute, with pay-as-you-go top-ups when a team exceeds it. That structure means the headline price is a floor, not a ceiling, and real cost tracks agent activity — Augment itself published that 22.5 percent of its users were consuming twenty times what they were paying, which is the context for its pricing changes. Enterprise is sales-quoted and is where SSO, OIDC and SCIM, customer-managed encryption keys, ISO 42001, multi-region compute, unlimited concurrency and volume discounts sit. Earlier free daily-request allowances have been withdrawn, and there is no clearly published free trial, which makes low-commitment evaluation awkward.
Security & Compliance
Sources
This page was written from 5 sources, 3 on domains other than augmentcode.com.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Raindrop
AI agent monitoring that catches silent production failures: Sentry for AI agents
GitLab Duo Agent Platform
Agentic AI across the whole GitLab DevSecOps lifecycle: planning, coding, code review, CI/CD and security agents under one governance model
CodeRabbit
AI code review and agentic change management for teams shipping human- and machine-written code
Qodo
Agentic code review and governance layer for teams shipping AI-generated code at scale