# Devin Desktop — AI IDE with a Kanban command center for running local and cloud coding agents; the renamed Windsurf. > Source: The Agents Index — https://theagentsindex.com/windsurf (structured, researched, re-verified) > Facts last verified: 2026-09-03 Devin Desktop is Cognition's AI IDE, the renamed successor to Windsurf, built around a Kanban Agent Command Center for local and cloud coding agents. Spaces group the sessions, pull requests, files and context for one task. | Fact | Value | | --- | --- | | Website | https://devin.ai/desktop | | Price | from $20/month | | Free tier | Yes | Free plan at $0/month, 1 member: light quota that resets daily and weekly, limited model availability (premium models crossed out), unlimited inline edits and unlimited Tab completions. No Devin Cloud agents, no Devin API, no DeepWiki, no Ask Devin, no sharing/collaboration, no concurrent-session allowance and no integrations (Slack, Teams, Linear, Jira, GitHub, GitLab, Bitbucket). App Deploys capped at 1 per day. When the quota runs out you wait for the next reset instead of buying extra usage. ## What it does **Writes code** · Yes — The in-editor agent authors new code and whole projects from a prompt. > You can create brand new projects with the agent! Click the "New Project" button to get started. > — https://docs.devin.ai/desktop/getting-started · Things to Try, "Generate a project with the agent" accordion **Debugs and fixes** · Partly — End-to-end debugging is claimed for the delegated cloud Devin agent, which is bundled with Pro, Max and Teams but absent from Free. > Devin handles complex tasks end to end — debugging, deployment, testing, and more. > — https://docs.devin.ai/desktop/devin · "What Devin does" section **Writes tests** · Partly — The only test-writing claim describes cloud Devin sessions delegated from a Space; no equivalent statement is made for the local agents. > an "Onboarding Flow Redesign" space might have one local agent session prototyping the UI and two cloud Devin sessions handling API changes and writing tests > — https://docs.devin.ai/desktop/spaces · Intro paragraph describing a Space **Reviews code** · Yes — Quick Review puts a second agent over the diff; the SWE-check model is free on every tier, GPT 5.5 and Opus 4.7 bill per token, and Enterprise admins must enable it. Devin Local agent only. > Quick Review runs an agentic code review on your local changes. When working with AI-generated code, Quick Review provides an independent second opinion > — https://docs.devin.ai/desktop/quick-review · Opening paragraph **Whole-codebase work** · Yes — Retrieval is delegated to the SWE-grep Fast Context subagent; Codemaps chart execution flow across a repository and Teams/Enterprise can index remote repos server-side. > Fast Context is a specialized subagent in Devin Desktop that retrieves relevant code from your codebase up to 20x faster than traditional agentic search. > — https://docs.devin.ai/desktop/context-awareness/fast-context · Opening paragraph **Ships it** · Partly — One-click deploys exist but only through the legacy Cascade agent, only to Netlify, in beta, and capped at 1 deploy/day on Free and 10 on Pro; agents otherwise stop at a pull request. > App Deploys run through Cascade tool calls and are unavailable with the Devin Local agent. > — https://docs.devin.ai/desktop/cascade/app-deploys · Overview section, final line ## How much it does unattended **Runs autonomously** · Partly — A task can be handed to a cloud Devin session that runs unattended on its own VM, while the local agent's default permission mode prompts before actions until Allow rules are set. > You can work on a plan with a local agent and, with a single click, send it to Devin for implementation. Devin spins up its own machine and gets to work while you keep coding locally > — https://docs.devin.ai/desktop/devin · "Delegating work to Devin" section **Multi-agent** · Yes — Local, cloud and third-party ACP agents run side by side in the Kanban Agent Command Center, Spaces share task context between them, and the local agent spawns its own subagents. > The Devin Local agent can spawn independent subagents to handle subtasks — either in the foreground or background. Subagents share tools and codebase context with the parent agent > — https://docs.devin.ai/desktop/devin-local · Key improvements → Subagents **Agent permissions** · Yes — Deny/Ask/Allow rules per action class, OS-level sandboxing with filesystem and domain filtering, and MCP tool calls prompting by default; Enterprise admins can enforce all of it. > Permissions can be scoped to file reads, file writes, command execution, HTTP fetches, and MCP tools. They can be configured at the project, user, or organization level. > — https://docs.devin.ai/desktop/devin-local · Differences → Permissions model ## Which models it runs on **Claude** · Partly — Claude is a premium model: the plan table marks premium models absent on Free and present from Pro upward; the model table lists Opus, Sonnet and Haiku variants. > Increased quotas, including access to OpenAI, Claude, and Gemini frontier models > — https://devin.ai/pricing · Individual plans, Pro plan card feature list **GPT** · Partly — Same gate as Claude: OpenAI frontier models come with Pro and above, while Free is marked "Limited model availability" with premium models crossed out. > Increased quotas, including access to OpenAI, Claude, and Gemini frontier models > — https://devin.ai/pricing · Individual plans, Pro plan card feature list **Open models** · Partly — Open-weight models are hosted by Cognition and included from Pro; the model cost table carries DeepSeek, Moonshot (Kimi), Z.ai (GLM) and Nvidia entries. You cannot point it at your own weights. > Free use of SWE 1.7 and leading open source models > — https://devin.ai/pricing · Individual plans, Pro plan card feature list **Model choice** · Yes — You pick the model per conversation, or hand it to the Adaptive router; Enterprise admins can restrict the selectable set by model or by provider. > You can easily switch between different models of your choosing. > — https://docs.devin.ai/desktop/models · AI Models page, above the model cost tables ## Where you use it **In your editor** · Yes — The product is itself the IDE: a VS Code-derived editor for macOS, Windows and Linux that imports VS Code and Cursor settings. Extensions cannot be installed from any marketplace. > Devin Desktop is a next-generation AI IDE built to keep you in the flow. > — https://docs.devin.ai/desktop/getting-started · Welcome to Devin Desktop, opening paragraph **On the command line** · Partly — Devin Desktop installs a devin-desktop shell command for opening files, and its local agent shares the harness with Devin CLI, but the terminal agent is a separate product. > Devin Local is our next-generation agent harness shared with Devin CLI. > — https://docs.devin.ai/desktop/devin-local · Opening paragraph **In your pipeline** · Partly — PR review automation over a GitHub org is an Admin Portal toggle and agent PRs surface in Spaces, but the IDE does not run as a CI job; Devin Review is a separate product line. > Enable PR review automation and description editing > — https://docs.devin.ai/desktop/guide-for-admins · 3.1 Admin Portal Overview → Feature Toggles → PR Reviews (GitHub Integration) **In a browser** · Partly — Devin Desktop is a downloaded desktop application; only the cloud Devin sessions it dispatches are also reachable in the app.devin.ai web app. > Devin sessions in Devin Desktop stay in sync with the web app > — https://docs.devin.ai/desktop/devin · "Working in a Devin session" section ## Whose machine it runs on **Self-hosted** · Partly — Enterprise-only, and the deployment docs describe Cognition hosting in a customer-isolated single-tenant VPC reached over AWS PrivateLink rather than customer-run software. > Deploy in your virtual private cloud (VPC) > — https://devin.ai/pricing · Compare Plans table, Security section, VPC row (Enterprise column only) **Open source** · No — Terms of service forbid decompiling or deriving the software, and no source repository or licence is published for the app. > reverse engineer, disassemble, decompile, decode, adapt, or otherwise attempt to derive or gain access to any software component of the Services > — https://cognition.com/pages/terms-of-service · Section 2.3 Restrictions, clause (ii) ## What it costs to run **How it meters** · Yes — A subscription (plus per-seat on Teams) buys a token-metered daily and weekly quota; overflow is billed at API list prices, and Enterprise contracts meter in Agent Compute Units. > Your budget is based on how many tokens the model uses for each request. The cost per token varies by model, and free models don't count against your quota at all. > — https://docs.devin.ai/desktop/accounts/quota · How quotas work **Free tier** · Yes — A $0 plan you download and sign into; no sales call, but no premium models, no cloud agents, no Devin API and a quota you must wait out when it runs down. > Light quota to code with agents > — https://devin.ai/pricing · Individual plans, Free plan card ($0) feature list **API access** · Partly — The Devin REST API (service users, RBAC, sessions, playbooks, audit logs) is crossed out for Free and ticked from Pro upward in the plan comparison. > The Devin API enables you to integrate Devin into your applications, automate workflows, and build powerful tools. > — https://docs.devin.ai/api-reference/overview · API Overview, opening paragraph **MCP server** · Partly — Devin Desktop is an MCP client that consumes servers with per-tool approval; the server another agent can query is the Devin platform's, at mcp.devin.ai, not the desktop session. > Beyond repository documentation and search, it gives any MCP-compatible AI agent or IDE full access to session management, playbooks, knowledge, and scheduling. > — https://docs.devin.ai/work-with-devin/devin-mcp · Opening paragraph **Bring your own key** · No — Inference is Cognition's; no customer LLM keys, and the closest thing to your own cloud is the Enterprise dedicated deployment with customer-managed keys. > Devin is a compound AI system and does not currently support third-party LLM API keys. > — https://docs.devin.ai/enterprise/deployment/overview · FAQs & Additional Information → "Can we use our own LLM API keys?" ## Buying it for a team **A company can buy it** · Yes — Teams and Enterprise are org purchases with centralized billing, an admin dashboard and priority support; Free, Pro and Max are single-member plans. > Our Teams and Enterprise plans unlock collaboration features such as team management, Single Sign-On (SSO), advanced analytics, and priority support. > — https://docs.devin.ai/desktop/accounts/teams-getting-started · Getting started with Teams and Enterprise, opening paragraph **Seat model** · Yes — Teams is a $80/month platform fee plus $40/month per full developer seat with unlimited flex seats and no published ceiling; Enterprise seat pricing is quoted. > /month for team plan + $40/mo per full dev seat > — https://devin.ai/pricing · Team plans, Teams plan card price cell (below $80) **Pooled budget** · No — Each paid full seat carries its own daily/weekly quota rather than drawing on a shared pool; only the older credit-based enterprise plans pooled add-on credits across the team. > Each full user includes their own generous quota and access to Devin Desktop > — https://devin.ai/pricing · Compare Plans table, Members row, Team column tooltip **Admin controls** · Yes — Admins cap terminal auto-execution, set command allow/deny lists, restrict models by model or provider, gate MCP, deploys and the local agent, and push MDM policies to machines. > Set the maximum auto-execution level for terminal commands across your organization > — https://docs.devin.ai/desktop/guide-for-admins · 3.1 Admin Portal Overview → Feature Toggles → Auto Run Terminal Commands (Beta) **Audit log** · Partly — An organization audit-log endpoint exists in the v3 API but requires an enterprise-level service user with ManageEnterpriseSettings, and queries are capped at 100-day windows. > List audit logs for the organization. > — https://docs.devin.ai/api-reference/v3/audit-logs/organizations-audit-logs · List Organization Audit Logs, endpoint description **Single sign-on** · Partly — SAML 2.0 and OIDC with SCIM provisioning, but Enterprise-only since the March 2026 repackaging; earlier Teams buyers of the SSO add-on keep it grandfathered. > New Teams plans do not include SSO — it is now an Enterprise-only feature. > — https://docs.devin.ai/desktop/accounts/quota · Other questions → "What happened to SSO for Teams?" ## What happens to your code **Opt out of training** · Partly — Opt-out lives in Data Controls and is available only on paid plans (admin-only on Teams), after which Zero Data Retention is enabled with model providers; Enterprise needs written consent to train. > By default, we may use your data for model training purposes to improve and enhance the Services. If you're on a paid plan, you can opt out at any time on the Data Controls settings page. > — https://docs.devin.ai/admin/security · Privacy & Intellectual Property → "How is your data used to improve Devin?" **Data residency** · Partly — No region choice is published for the commercial service, which runs in Cognition's AWS environment; dedicated Enterprise deployments keep data in the customer's tenant and the federal deployment runs on AWS GovCloud. > For Enterprise customers with dedicated deployments, all customer data is stored within the customer's tenant. > — https://docs.devin.ai/enterprise/security-access/security/enterprise-security · Privacy & Intellectual Property → "How does Cognition process data accessed by Devin?" **Getting out** · Partly — Unspent extra-usage balance is refundable on request and admins can export analytics reports; the included plan quota resets daily and weekly rather than accumulating, and no agent-conversation export is documented. > After conversion, you can request a refund of your unspent extra usage balance at any time through support. > — https://docs.devin.ai/desktop/accounts/quota · Add-on credits & extra usage conversion → "Can I get a refund for my unspent add-on credits?" **Certifications** · Partly — SOC 2 Type II since September 2024, and Devin Desktop is marked available at FedRAMP High, IL4/IL5 and ITAR with Zero Data Retention in federal deployments; the docs claim no ISO 27001 or GDPR position. > Cognition has been SOC 2 Type II certified since September 2024. > — https://docs.devin.ai/enterprise/security-access/security/enterprise-security · Security → Third-Party Audits and Certification ## Provenance 2 of 37 sources behind the facts above are independent of the vendor. ## Coverage elsewhere Pages about this product that the vendor does not control. The label states how independent each one is. | Source | Kind | Read | | --- | --- | --- | | cloudzero.com | Independent source | 2026-07-21 | | developers.googleblog.com | Independent source | 2026-07-24 | | github.com | Platform record | 2026-07-24 | | infoworld.com | Independent source | undated | ## What changed - 2026-09-03 — Capabilities · Commercial terms · Free tier - 2026-08-31 — Capabilities · Commercial terms · Free tier · Price - 2026-08-22 — One-liner - 2026-08-19 — Summary ## Correcting this record Anything marked "Not established" is a question we have not answered — it is not a No. If you represent Devin Desktop, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth. Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.