What Augment Code does
Augment Code is an AI software-development platform with IDE agents, the Auggie CLI, rules, codebase retrieval, and an MCP context engine for multi-repository work.
Augment Code centers on repository context rather than a blank code chat. Its agents can retrieve relationships across large codebases, use workspace rules, and operate in IDE or CLI workflows. A separate MCP context service can expose code and documentation retrieval to compatible coding clients, including local or remote multi-repository modes.
Current credit-based individual tiers are listed at $20 for Indie, $60 for Standard, and $200 for Max, with a trial allocation and custom Enterprise terms; credits measure differing model and task costs. Pricing and included credits have changed several times, so teams should use the live dashboard and estimate representative refactors, tests, and context queries rather than assuming each request costs the same.
Augment says it does not train on proprietary customer data and publishes SOC 2 Type II and AI-management security claims. Indexing broad repositories still creates exposure risk, and an agent can introduce insecure or incorrect code. Scope repository access, exclude secrets, inspect generated commands and diffs, pin tool permissions, run independent tests and scanners, and keep protected branches and human review.
How Augment Code works
A developer installs Augment in a supported IDE or uses the Auggie CLI, authorizes selected repositories or context sources, and supplies a request plus optional rules. The Context Engine semantically indexes and retrieves relevant code and documentation, while the selected agent model plans, reads, edits, runs permitted tools, and returns a patch or explanation. Rules can be local, user-level, repository-level, or hierarchical. Every change still requires diff review, tests, security checks, and human approval before merge.
How to set up Augment Code
Choose an isolated pilot repository
Start with a non-sensitive codebase, named maintainers, representative tasks, required tests, and a clear rule against direct production changes.
Install the appropriate surface
Use the supported IDE extension or Auggie CLI and authenticate only the repositories and documentation needed for the pilot.
Add precise project rules
Record versions, architecture boundaries, commands, style, security requirements, and prohibited files in reviewable workspace rules.
Evaluate real engineering tasks
Measure retrieval relevance, patch correctness, test quality, credit use, latency, secret handling, and performance on ambiguous requests.
Enforce the normal delivery gate
Review every diff and command, run tests and security scans outside the agent, require approvals, and monitor permissions and spend.
Augment Code FAQs
How is Augment Code billed?
Current individual plans use monthly credit pools whose consumption varies by task and model. Enterprise and team arrangements can differ.
Does it work outside an IDE?
Yes. Auggie provides a CLI workflow, and Context Engine MCP can supply retrieval to supported MCP clients.
Does Augment train on private code?
Augment states that it does not train on customers' proprietary data. Organizations should still review retention, subprocessors, and contract terms.
Are generated changes safe to merge?
Not automatically. Agents can misunderstand requirements, introduce vulnerabilities, or write weak tests. Human diff review and independent validation remain required.
What changed in 2026?
Official documentation notes that code completions and Next Edit were deprecated for several individual plans on March 31, 2026; Agent and Chat remain central.
Listing reviewed 2026-07-27. Product details and pricing can change; verify important terms on the provider's website.
Related Coding AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.