● aa index
—
Pair not measured
Starting at
from $0
Free plan with limits; session fleet - from Pro $20/mo
Surfaces
6+
cloud · desktop · cli · slack-teams · api
Latest release
29.06
From the card's track record
Quick start · Web (cloud)
$ → app.devin.ai▌benchmarks and competitors - analysis in 3.0 ↓
1.0
Should you get it
Use it if you need
- ✓batches of delegable tasks: bug fixes, migrations, small features - in parallel in the cloud
- ✓teams with a review process: Devin’s output - a PR approved by a human
- ✓delegation from Slack/Teams, Jira, and Linear - the developer’s machine stays free
- ✓enterprise with VPC deployment, roles, and SSO requirements
Choose it for delegating batches of tasks to the cloud with review at the end; skip it as your only daily agent - and if you are in Russia.
Pass by if needed
- ✗daily hands-on work in your own repo - the terminal-agent category (Claude Code, Codex)
- ✗predictable budget: message pricing varies, with no official formula
- ✗working from Russia: directly prohibited by the terms of service
- ✗open-source bet: the core is closed, with no product repository
Free plan $0 with limits; fleet starts at Pro $20/mo, Max $200/mo, Teams from $80/mo. Prohibited from Russia under the terms of service.
2.0
Installation by surface
Devin Cloud (app.devin.ai)→
Main surface: the task goes to a cloud VM with a browser, terminal, and editor, producing a PR. Windows sessions are available separately and cost about 9% more than Linux.
→Connect a GitHub repository and describe the task with acceptance criteria
>_Devin Desktop→
Desktop IDE (announced 06/02/2026): manages local and cloud sessions from one window; model selection including open source (Kimi K2.7, GLM 5.2).
>_CLI→
Terminal agent with cloud handoff: start locally - continue in a cloud session. The model is selectable (Adaptive router, opus, sonnet, codex, gemini, swe-1-6).
$ curl -fsSL https://cli.devin.ai/install.sh | bash$ devin>_Slack and Microsoft Teams→
Tag @Devin in a thread - the task goes to a cloud session, and the response returns to the thread. The main delegation path for teams.
→Connect the integration in Settings: Integrations, then Slack / Teams
→Tag @Devin in a thread with a task description
>_API→
Programmatic session launches: queue tasks from your own systems (tickets, cron jobs, CI) without a web interface.
>_Devin in Windsurf→
Inside Cognition's acquired Windsurf editor (post 04/15/2026): the agent next to the tab with inline edits, one subscription.
What you need before starting
Requires a devin.ai account and connected GitHub. Payment by foreign card; the service is prohibited from Russia under its terms.
● Availability from Russia
Russia is explicitly named among the restricted jurisdictions in the service terms - stricter than Anthropic's "not in the country list." There is no local equivalent of a cloud fleet; the closest option is SourceCraft Code Assistant (Yandex).
3.0
Rankings and benchmarks
No Devin+model pair on the leaderboards · card reference points
SWE-bench (unassisted)
13.86% · 03.2024 · the only official measurement of Devin itself; no recent ones
AA Index / Terminal-Bench
not measured · 07.2026 · no Devin pair in leaderboards
FrontierCode Diamond (model)
13.4% · 06.2026 · Cognition's own benchmark, measures models (best - Opus 4.8), not Devin
Who is measured and how it is calculated - AI agent ranking
● agent ≠ model
Devin is not measured in AA Index or Terminal-Bench: a cloud fleet does not fit into terminal-agent benchmark constraints. The table is a class map for orientation; Devin's own figures are in the rows below.
4.0
What it can do: autonomy, techniques, character
Autonomy axis
Devin's autonomy axis goes further than anyone else's: not an agent in your terminal, but a fleet of sessions in someone else's cloud. The price is acceptance: everything returned is checked by a human.
Devin· covers the entire axis
01Task setup
task from the web, Slack/Teams, Jira, or API - the developer's machine stays free
02Autonomous session
its own VM with browser, terminal, and editor; Devin plans and executes on its own
03Fleet
parallel sessions: a batch of tasks goes to the cloud at once
04Acceptance
output - PR; results are not accepted without human review
Asks at every stepWorks while your laptop is closed
Project character · releases up, incidents down
2026-06-29
release
Devin Fusion - multi-model router - Work is split between a frontier model and a cheaper parallel one: $2.38/task versus $3.00 for pure Fable 5 on FrontierCode Extended.
2026-06-24
release
Kimi K2.7 and GLM 5.2 in Desktop and CLI - Open-source models as an option; free through July 5 on Pro/Max/Teams.
2026-06-20
policy
Hard session caps, per-session ACU limit lowered to 1 - A response to unpredictable spending from long sessions; ACU limit audit in the June release notes.
2026-06-08
release
FrontierCode - proprietary PR mergeability benchmark - Measures models (correctness, test quality, scope discipline); affiliated with Cognition.
2026-06-02
release
Devin Desktop - fleet management from one window
2026-05-27
release
Round over $1B at a $26B valuation, $492M run-rate
5.0
Pricing, limits, and access from Russia
Plans · $/mo
free
Free
limited usage
$20
Pro
daily + weekly quota; entry into the fleet
$200
Max
weekly quota without a daily cap
$80
Teams
team minimum + $40 for a full seat
Plans
Free $0 · Pro $20/mo · Max $200/mo · Teams from $80/mo
Pro $20/mo: daily and weekly session quotas, CLI, and Desktop. Max $200/mo: weekly quota with no daily cap. Teams: at least $80/mo plus $40 per full seat; flex seats are free and use the shared pool. Enterprise is priced in ACUs by contract. Purchased credits do not expire.
Devin pricing
2026-07-06
Free access
Yes: Free plan with limits
Free provides limited Devin use, limited models, Devin Review, and DeepWiki. A serious cloud session fleet starts with Pro $20/mo.
Devin pricing
2026-07-06
Limits
Daily + weekly quotas; Enterprise - ACU
Pro and Teams: daily plus weekly quota; Max: weekly only. Per-message pricing varies by model, task size, and complexity - there is no official formula. Enterprise uses ACU: agent actions plus VM time; Windows sessions cost about 9% more than Linux. Hard session caps were introduced in June 2026.
Devin docs: usage
2026-07-06
Local alternatives
No direct equivalent; the closest is SourceCraft (Yandex)
There is no direct Russian equivalent of a cloud fleet of agents. The closest local option is SourceCraft Code Assistant (Yandex): an agent mode "from task to pull request," billed in rubles.
SourceCraft
2026-07-06
● limits expire
All vendors in 2026 rebuilt billing around token credits; limit figures change every week. Each line above has a verification date - do not trust figures without a date, including ours.
6.0
Risks: privacy and control
Code privacy
No code training by default
By default, Cognition does not train models on customer data and code; SOC 2 Type II since September 2024. Enterprise: deployment in the customer's VPC, SSO, roles, IP lists. Output code belongs to the customer.
Devin docs: security
2026-07-06
Availability from Russia
Prohibited by the terms of service
Russia is explicitly listed among ToS restricted jurisdictions (along with Belarus, Iran, and others). This is stricter than Anthropic: not "not on the list of supported countries," but an explicit prohibition.
Terms of Service
2026-07-06
Open source
No, the core is closed
There is no public product repository: github.com/CognitionAI has only supporting tools (blockdiff, deepwiki, SWE-bench results).
GitHub CognitionAI
2026-07-06
Company
Cognition: round over $1B at $26B (05/2026)
$492M run-rate revenue; Devin commits 89% of Cognition's internal code. In July 2025, Cognition acquired Windsurf (~$250M); the product is alive and folded into unified pricing. Customers: Citi, Goldman Sachs, Mercedes-Benz.
Cognition blog
2026-07-06
7.0
Devin or Claude Code - cloud versus terminal
Devin - a cloud fleet: assign a batch of tasks and accept PRs. Claude Code - terminal-first: depth in your own repo here and now. The choice is "where the work lives," not "who is smarter."
Devin
Claude CodeCardfleet of parallel cloud sessions; batch delegation
Strength
depth in a large codebase, careful edits
not measured in AA Index or Terminal-Bench; own SWE-bench 13.86% (03.2024)
Benchmarks
Terminal-Bench 78.9% with Opus 4.8 (06.2026)
Free $0; Pro $20/mo; message price varies - no formula
Economics
from $17/mo; limits in 5-hour windows + weekly
Cloud, Desktop, CLI, Slack/Teams, API, Windsurf
Surfaces
14: CLI, IDE, desktop, web, iOS, CI, Slack, SDK
Fusion router; CLI and Desktop selection down to open source
Models
Claude only (API/Bedrock/Vertex/Foundry)
core is closed, no product repository
Openness
no, proprietary CLI
- →Devin is a second layer: a fleet for delegable batches on top of a terminal agent for daily work.
Other paths
Devin or Codexcloud tasks already in a ChatGPT subscription: $0 entry
Devin or Windsurftheir editor too: if you need an IDE, not a fleet
Devin or Copilot Coding Agentdelegation directly in GitHub, if the team lives there
Frequently asked questions
Devin or Claude Code - which should you choose?
Different workflows. Devin is a cloud fleet: batches of tasks are submitted from Slack or tickets, producing PRs for review; the developer machine stays free. Claude Code is a terminal-first agent for deep work in your own repo, here and now. In 2026, pros more often keep a terminal agent as the foundation and a fleet as a second layer for delegable batches.
How much does Devin cost?
Free $0 with limits; Pro $20/mo (daily and weekly quotas); Max $200/mo (no daily cap); Teams from $80/mo plus $40 per full seat. Enterprise is priced in ACUs by contract. Caution: per-message pricing varies by model and task complexity - there is no official formula (verified 07/06/2026).
Does Devin work in Russia?
No. Russia is explicitly listed among the restricted jurisdictions in the terms of service - this is stricter than Anthropic's "not on the list of supported countries." This card does not provide workarounds. The closest local option is Yandex's SourceCraft Code Assistant, but it is not a cloud fleet of agents.
Is Devin free?
Partly: the Free plan provides limited use, Devin Review, and DeepWiki. A serious cloud session fleet starts with Pro $20/mo.
What is ACU in Devin?
Agent Compute Unit - the billing unit for Enterprise contracts: it accounts for agent actions (planning, browser, code execution) and VM time. There is no public ACU price - "rate from the order form"; the common formula "1 ACU = 15 minutes" is absent from the current docs.
What is correct: Devin, devin, or devin ai?
The canonical name is Devin (a Cognition product). In Russian, it is written as "devin" and "devin ai" - this card is found by all variants. Do not confuse it with namesake Devin packages in npm: the product has no official npm package.
What happened to Windsurf after Cognition acquired it?
It is alive: Cognition acquired Windsurf in July 2025 (~$250M) and folded it into unified pricing - "Devin in Windsurf" has worked since April 2026. The editor remains a separate surface: if you need an IDE rather than a fleet, look at Windsurf.
Watch and read on the topic
Entity · devin
Facts · 13, each with a source and date
Card editorial review · 2026-07-06
Facts verified · 2026-07-06
Where to go next
Rating updates
New cards, price and limit changes, ranking shifts - by email, no spam.