At a glance
- Category
- Coding agents
- Tags
- HIPAA, ISO 42001
For developers
- Works with
- Slack, VS Code, JetBrains, GitHub, VSCode, Terminal and 18 more
Cosmos, an agent platform, is now the product on sale
Augment started as a coding assistant, and the site now leads with something larger. The homepage sells Cosmos as "the platform for building an always-on software delivery system that moves work from spec to verification", made of loops: code review, ticket to PR, vulnerability remediation and incident response. Each loop is run by Experts, prebuilt agents such as PR Author, Deep Reviewer, Risk Analyzer and Verifier that a team can fork and tune, and the Cosmos page lists twelve of them. Work starts from triggers in GitHub, GitLab, Slack, Linear and Jira, runs in isolated VMs or on a self-hosted daemon, and pauses at checkpoints where a person approves.
The Auggie CLI is the second product: an npm install with an interactive mode and a --print mode for CI, plus sub agents, parallel agents, resumable sessions and per-environment tool permissions. The VS Code and JetBrains extensions still ship releases on the changelog, but the pricing page lists only Cosmos, CLI access and MCP tools on each plan, and the docs table for the Business plan names "Cosmos and Auggie CLI" as the products included.
Two flat team plans, each priced at its own usage allowance
The pricing page has three plans. Standard is $20 a month and Business $100 a month, both "flat, no per-seat charge" for up to 50 seats, and each includes usage equal to its price: $20 or $100 a month across LLM, Context Engine and compute. Enterprise is Custom, with unlimited users, "Volume-based annual discounts", SSO, OIDC and SCIM, CMEK, multi-region compute and unlimited concurrent Cosmos sessions. Standard and Business share standard-size compute and a cap of 50 concurrent sessions.
The included usage is pooled: "Heavy users and occasional users draw from the same shared balance", per the FAQ. A team of 50 on Standard therefore shares $20 of model spend a month, so the seat count is a ceiling on access, and the real bill is set by top-ups. Top-ups bought outside the plan stay valid for 12 months. No monthly and annual toggle appears on the page; the only annual term is the Enterprise discount. The FAQ mentions trials only to say they get community support and no SLA, and publishes no trial length or allowance.
Overage is model list price plus 40%, and compute by the hour
The token pricing docs explain what the dollars buy. Every request is billed as LLM tokens "at the model provider's public API list price", a "flat 40% service fee on LLM usage", and Cosmos compute at $0.19 an hour "billed in 5-minute increments (rounded up)". Once the included amount is spent, billing "continues automatically on pay-as-you-go" with "no minimum top-up amount". The vendor's sample Business month spends $70 on tokens, $28 in fees and $2 on compute.
The same page carries a per-million-token rate table across Claude, GPT, Gemini, Grok, DeepSeek, Kimi and GLM models, plus Prism, a router whose rates are "Variable (routed)". Model choice is therefore the main cost lever, and the spread is wide: the table runs from $0.10 to $10.00 per million input tokens. The Cosmos page offers spend caps per automation, per user or per group. The docs page describes only the Business plan and does not mention Standard.
Harness bought Cosmos, the CLI and the Context Engine in October
The homepage banner reads "Augment Code Joins Harness", and every page footer now says "© 2026 Harness, Inc."
The wording is "select" assets, and the IDE extensions are not among those it names. The augmentcode.com pages still sell under the Augment name and prices, so the brand, billing entity and product roadmap are all in transition. The changelog has no entry after October 7.
Security claims are published, with the strongest ones on Enterprise
The security page lists SOC 2 Type II as attested and ISO/IEC 42001 as certified, states GDPR, CCPA and HIPAA compliance, and says a BAA is available. It says "We never train on our customer's proprietary data", and the pricing FAQ puts that in contract terms for every paid plan. Reports are requested through a trust center rather than published.
The split by tier matters. The Standard and Business cards list SOC 2 Type II and "No AI training allowed"; SSO, OIDC and SCIM, CMEK and ISO 42001 appear on the Enterprise card only. Support also differs: Standard and Business get community support plus tickets handled under the SLA, while Enterprise gets "Full SLA coverage with dedicated support".
The Context Engine is sold to other agents over MCP
Augment's indexing layer is available outside its own tools. The Context Engine MCP docs offer a local server, which runs the Auggie CLI over stdio and indexes the working directory as it changes, and a remote server hosted by Augment that indexes the default branches of repos connected through its GitHub App. Quickstarts cover Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Kiro and others.
These queries bill under the same token pricing, and the docs put the cost at "roughly $0.03–$0.06 per query" on average. A team that already uses another coding agent can buy the codebase index alone, but each query draws on the same pooled balance as the rest of the plan.
The site disagrees with itself on service accounts and models
The CLI page describes service accounts with dedicated API tokens and ends the section "Enterprise only." The September 17 entry on the changelog says "Service accounts are now available on Enterprise and team plans." The same CLI page offers "Claude, GPT-4, or bring your own", while the rate table lists far newer model families and no GPT-4. The docs and changelog are the more recent of each pair.
It fits engineering teams that want agents running review, ticket and incident work on a schedule with approval gates, and that can watch a metered bill. A solo developer wanting inline completions in an editor is buying the wrong thing: the plans are priced and described around Cosmos and the CLI, and those products now belong to Harness's software delivery platform.
Augment Code price and terms
Terms
- Free to start
- PartlyPartlyThe pricing FAQ refers to trials but does not publish their duration, allowance, or eligibility
What needs a higher plan
- Enterprise
- Public API
For AI agents
- MCP server
- Yes
- API
- PartlyEnterprise

Pricing page, captured
augmentcode.com/pricing, checked
All plans and limits
| Plan | Price | Main limits |
|---|---|---|
| Standard | $20 a month | 50 concurrent sessions; Included usage: $20/mo of usage across LLM, Context Engine & compute; Up to 50 seats included |
| Business | $100 a month | 50 concurrent sessions; Included usage: $100/mo of usage across LLM, Context Engine & compute; Up to 50 seats included |
| Enterprise | Price on request | Unlimited concurrent sessions; Bespoke usage limits; Unlimited users |
Read from augmentcode.com/pricing.
What we checked on Augment Code
What the price counts
A flat team price with no per-seat charge for up to 50 seats, which includes a dollar usage allowance; usage beyond that is metered at model list price plus a 40% fee and compute time.
One flat price for your whole team, up to 50 seats with $20 of usage included every month.
augmentcode.com/pricingChecked
Free tier
You can buy the $20 plan self-serve with no sales call, and the support policy refers to trials, but the pricing page publishes no standing free plan or trial terms.
Trials and beta use: Community support only; our SLA does not apply to trials or beta usage.
augmentcode.com/pricingChecked
Public API
A public Analytics API with service-account tokens exists but is preview and Enterprise-only; everyone else drives the product through the CLI and SDKs instead.
The Analytics API is currently in preview and is available exclusively to Enterprise customers.
MCP server
The Context Engine is exposed to other agents over MCP, locally via the CLI or as a hosted HTTP server, and a docs MCP server is published too.
Remote Server (Best for Adding/Understanding Code): Connect to Augment-hosted Context Engine over HTTP
Open source
Context Connectors is published as an open-source library on the Context Engine SDK, while the agent, Cosmos platform and CLI are distributed as closed products.
An open-source library built on the Context Engine SDK that makes diverse sources searchable across agents and apps
Augment Code alternatives
Augment Code and Hyper
Augment Code beside its alternatives: the price on each one's site and the day it was checked Tool Price Checked Augment Code From $20 a month Hyper Free, then $20 a month
Questions about Augment Code
- How much does Augment Code cost?
- Augment Code costs $20 a month. From augmentcode.com/pricing, checked 5 November 2026.
- Does Augment Code have an API or an MCP server?
- It has an API on Enterprise. It has an MCP server.
- Is Augment Code open source, and can you host it yourself?
- Part of it is open source.
- What have we checked on Augment Code?
- 5 facts, under these headings: What it costs to start, how you build with it.
- How do we check Augment Code?
- We hold 5 facts on Augment Code, quoted from 7 pages. Facts and price last checked 5 November 2026.
Sources and changes
Last changed
Similar to Augment Code
- ERIUS PHONERent isolated virtual Android phones by the month so an AI agent can install apps, tap through them and place calls.
- TinyFishWeb search, page fetch, cloud browsers and a goal-driven web agent for AI agents, with search and fetch free.
- UI DiscoveryA library of real websites, iOS apps and dashboards with each one's palette, fonts and original build prompt.
- ClineFlowGit-native project memory and handoffs for coding agents.
From our research
- Daytona vs E2B vs Modal: which sandbox for agent code?Daytona, E2B and Modal compared on sandbox pricing, concurrency, GPU support and open-source status, plus why Daytona's flagship repo is now archived.
- ERIUS PHONE review (2026): pricing, pros and consERIUS PHONE rents AI agents real cloud Android phones they drive by API, priced monthly per phone in euros, but it offers no webhooks.
- Hyper review (2026): an autonomous coding agent that ships verified pull requestsHyper plans, tests and delivers pull requests to your GitHub repository from an isolated cloud sandbox, with a verification check on each one and a hard spending cap on every run.
Is this yours?
Claim it free, then see who reads this page and carry the verified owner mark. You confirm it with an email at augmentcode.com.
Report an error if a fact is wrong.