Devin
Devin is Cognition Labs' cloud-based autonomous AI software engineer - it runs in a sandboxed environment with its own browser, terminal, and code editor and is built to complete full engineering tasks end-to-end rather than assisting in an IDE. Devin supports SSO, on-prem deployment, code review, and deploy integration. Pricing starts at $500/mo per subscription, targeting teams rather than individual developers.
Overview
Cloud-based autonomous AI software engineer that works in a sandboxed environment with a browser, terminal, and code editor to complete full engineering tasks.
Compare Devin side-by-side with other CLI, IDE, cloud, and framework agents on pricing, model coverage, and release cadence.
Best LLMs for this agent
Cognition routes Devin to Claude Sonnet 4.6 as a documented default for autonomous task execution; Devin manages model choice internally.
Claude Opus 4.6 is the managed option Cognition uses when Devin needs stronger reasoning for end-to-end engineering tasks.
Pricing and Provider Comparison
Hosted agents expose plan pricing. Bring-your-own-key agents pass the main cost through the model provider route you configure.
Agent access tiers
| Tier | Price | Includes |
|---|---|---|
| Starter | Free | 10 compute hrs, 1 parallel task |
| Pro | $30/mo | 100 compute hrs, 5 parallel tasks, 2M token context |
| Core | $20/mo | Base access + pay $2.25/ACU (~15 min active work each) |
| Team | $500/mo | 250 ACUs included, API access, Slack, IDE extension, unlimited seats |
| Enterprise | Custom | VPC deployment, custom integrations, dedicated support |
Featured Model Specs
| Model | Lab | Released | Context | Reasoning | Openness |
|---|---|---|---|---|---|
| Claude Sonnet 4.6 | Anthropic | 2026-02-17 | 1m | Yes | Proprietary |
| Claude Opus 4.6 | Anthropic | 2026-02-05 | 1m | Yes | Proprietary |
Model Compatibility Matrix
| Model | Lab | Default | Access tier | Notes |
|---|---|---|---|---|
| Claude Sonnet 4.6 | Anthropic | No | Core | Used internally for coding tasks |
| Claude Opus 4.6 | Anthropic | No | Core | Used internally for complex reasoning |
Underlying Models
Compare to Similar Harnesses
FAQ
Which LLM does Devin use?
Cognition routes Devin to Claude Sonnet 4.6 and Claude Opus 4.6 as documented defaults for autonomous task execution. Devin manages model selection internally; you don't bring your own key.
How much does Devin cost?
Devin subscriptions start at $500/mo per seat. Enterprise pricing with SSO and on-prem deployment is available on request.
Does Devin work in VS Code?
No. Devin is a hosted environment - it operates from its own sandboxed cloud workstation with browser, terminal, and editor access. You interact via the Devin web UI, Slack, or CLI hooks.
Is Devin autonomous?
Yes. Devin's design point is end-to-end task execution: it plans, runs commands, edits multi-file changes, opens PRs, and iterates against test feedback without per-step prompting.
Is Devin open source?
No. Devin is a closed-source commercial product from Cognition Labs.