Augment Code review

Augment Code builds Cosmos, a software agent orchestration platform for running workflows across the software development lifecycle. It sits above code-hosting, ticketing, chat and deployment tools: agents take work from a trigger, use codebase context and tools, then hand results to another workflow or a human reviewer.

Cosmos includes prebuilt Experts for tasks such as ticket triage, PR authoring, code review and testing. Teams can also fork or build Experts, each with its own environment, capabilities and memory. The vendor describes loops that connect ticket, code, review, deploy and monitor stages, with human approval checkpoints for consequential actions.

This is not just an editor extension or a model API. It is a platform for coordinating agents, with managed cloud sandboxes, execution on a team’s own machines, or a mix of both. Augment presents codebase understanding through its Context Engine, which gives agents context across an organization’s repositories.

The homepage names CLI access, MCP and Native Tools, plus typed webhooks for triggers from the rest of a stack. Native triggers are listed for GitHub, GitLab, Slack, Linear, Jira and schedules. The product can run agents in managed cloud sandboxes, on a team’s own machines, or across both environments.

Augment describes building workflows in natural language: an Advisor proposes the Experts, environment and automation, then a person approves the plan. The Experts run the workflow and can pause for human review before resuming. This is the main integration model described on the homepage, rather than a documented REST API or language-specific SDK.

The homepage links to product documentation, but it does not name SDKs or programming languages. Confirm SDK support, API style, authentication, versioning and webhook behavior in the vendor’s docs before designing an integration. CLI, MCP, Native Tools and typed webhooks are the integration surfaces the homepage explicitly names.

There is no free plan or free trial listed. Standard costs $20 per month per team, up to 50 seats, and includes $20 per month of usage across LLM, Context Engine and compute. Business costs $100 per month per team, also up to 50 seats, and includes $100 of that usage. Both include CLI access, MCP and Native Tools, 50 concurrent sessions, standard compute, SOC 2 Type II and no AI training on data.

Those fees include a usage allowance, not unlimited usage. The pricing page lists pay-as-you-go usage top-ups, valid for 12 months after purchase, and a 40% service fee on LLM usage; compute has no service fee. The published pricing does not state top-up rates or a hard usage ceiling, so model expected LLM and compute consumption with Augment before rollout.

Enterprise pricing is custom, with custom usage and top-up limits, unlimited users and concurrent sessions, multi-region compute, and security options including SSO, OIDC, SCIM and CMEK. Both listed team plans are monthly, with no annual discount shown.

The vendor does not publish a software licence or source-availability terms on the vendor's site reviewed here. Ask Augment whether any Cosmos components are open source and what licence applies before building around them. The product is not described as a self-hosted Cosmos server: Augment offers managed cloud sandboxes, execution on your own machines for code that cannot leave the building, or a mix of the two. Running agents on your machines is not the same as self-hosting the orchestration platform.

The vendor does not publish a data-export path or a migration process. Cosmos workflows, Experts, shared memory and codebase context may become part of a team’s operating setup, so ask what can be exported and in what format. Also confirm how workflow definitions, run history and memory behave if you stop using the service. The pricing lists no free plan.

Standard and Business list 50 concurrent sessions and standard compute, with teams capped at 50 seats. Enterprise offers unlimited users and concurrent sessions, custom usage and top-up limits, and multi-region compute. The pricing does not publish request-rate limits, uptime commitments or detailed scaling behavior; ask for those figures and incident-response terms before making Cosmos part of a critical delivery path.

Standard and Business list SOC 2 Type II and no AI training on data. Enterprise adds SSO, OIDC, SCIM, CMEK and ISO 42001. Confirm which data each workflow sends to models and tools, where it is stored, and how retention and export work.

Reddit users discussing an earlier credit-based pricing change reported high consumption, cancellations and dissatisfaction with support. Those comments are not a measurement of the current team plans, but they make usage accounting and support escalation worth checking. The current pricing lists a 40% service fee on LLM usage and pay-as-you-go top-ups.