Augment Code
Augment Code's Cosmos is an agent orchestration platform for running automated software delivery loops across the SDLC with human checkpoints.
At a Glance
Engagement
Available On
Alternatives
Updated Nov 2026
About Augment Code
Augment Code makes Cosmos, an agent orchestration platform for building what it calls a software factory: an always-on delivery system that moves work from spec to verification. The homepage notes that Augment Code has joined Harness, and the site footer carries Harness, Inc. copyright. Teams describe the loop they want in natural language, approve a plan, and the loop goes live.
What It Is
Cosmos runs software agents at scale, supplying the context, tools, and feedback loops they need. Work is organized into loops such as code review, ticket to PR, vulnerability remediation, incident response, test coverage, migrations, and onboarding. Agents handle the repetitive middle of the SDLC while engineers hold the quality gates.
How the Loops and Experts Work
Cosmos ships with Experts for each lifecycle stage. A Work Dispatcher triages tickets, a PR Author takes a task or ticket through review, CI, and merge, a Code Review Fleet (Pair Review, Deep Code Review, PR Risk Analysis) posts inline findings, and a Tester exercises changes end to end with screenshots. Experts are reusable templates with their own environment, capabilities, and memory. Users can fork them, build their own, or start from a library of 62 workflows with node diagrams and prompts. An Advisor expert helps design and wire up new loops.
Platform and Deployment
Triggers include GitHub, GitLab, Slack, Linear, Jira, schedules, and typed webhooks. Agents run in managed cloud sandboxes, on laptops, dev VMs, or the customer's own cloud (AWS, GCP). Documented platform pieces include an Expert Registry, human-in-the-loop escalation, shared organization memory, a Context Engine for codebase understanding, and a shared file system. The model picker offers auto-routing across Claude and Gemini models or single-model selection. The homepage states results such as 66% faster time to merge and 2-3x engineering throughput for specific loops; these are vendor claims.
Community Discussions
Be the first to start a conversation about Augment Code
Share your experience with Augment Code, ask questions, or help others learn from your insights.
Pricing
Standard
Flat price for whole team, up to 50 seats, with $20 of usage included every month.
- $20/mo of usage across LLM, Context Engine & compute
- Up to 50 seats included
- Cosmos
- CLI access
- MCP & Native Tools
- Top-ups, pay as you go
- SOC 2 Type II
- No AI training allowed
- Usage billed in dollars: LLM at provider's public API list price plus flat 40% service fee on LLM usage (no fee on compute)
Business
Flat price for whole team, up to 50 seats, with $100 of usage included every month.
- $100/mo of usage across LLM, Context Engine & compute
- Up to 50 seats included
- Cosmos
- CLI access
- MCP & Native Tools
- Top-ups, pay as you go
- SOC 2 Type II
- No AI training allowed
- Usage billed in dollars: LLM at provider's public API list price plus flat 40% service fee on LLM usage (no fee on compute)
Enterprise
For enterprise teams with high volume, security, or support needs.
- Cosmos
- Custom user pricing
- Bespoke usage limits
- Volume-based annual discounts
- SSO, OIDC, & SCIM support
- SOC 2 & Security Reports
- CMEK & ISO 42001 compliance
- Dedicated support
- No AI training allowed
Capabilities
Key Features
- Agent orchestration platform (Cosmos) for software delivery loops
- Code review, ticket-to-PR, vulnerability remediation, and incident response loops
- Reusable Experts with their own environment, capabilities, and memory
- Library of 62 agent workflows
- Advisor expert for designing loops in natural language
- Triggers from GitHub, GitLab, Slack, Linear, Jira, schedules, and webhooks
- Managed cloud sandboxes, local machines, or customer cloud execution
- Human-in-the-loop approval checkpoints
- Shared memory across runs, agents, and teams
- Context Engine for codebase understanding
- Model routing across Claude and Gemini models
- CLI and cloud agents
