GILD verdict
Choose Claude Code if engineers want to steer the agent and approve its actions as it works.
Choose Devin if a queue of scoped tickets should come back as pull requests without an engineer in the session. Claude Code comes inside a Claude Team seat, and teams that run both can hand longer tasks from Claude Code to Devin's cloud sessions.
How much of the work an engineer watches decides this, since reviewers merge the two agents' work at similar rates. In an academic study presented at MSR 2026, reviewers merged 72.6% of Claude Code's public GitHub pull requests and 68.0% of Devin's over the same 11 weeks of 2025. Devin takes tasks from Slack, Linear, Jira, or GitHub, with a terminal CLI alongside. Claude Code starts in the terminal, an IDE, or the desktop app with permission prompts, and adds cloud sessions on the web.
What teams report
Failures and incidents 1
Answer.AIOn Devin
Answer.AI gave Devin 20 real tasks over a month and recorded 14 failures, 3 successes and 3 inconclusive results, then finished failed work step by step in a developer-driven editor (Jan 8, 2025).
“In contrast to Devin, we found workflows where developers drive more (like Cursor) avoid most issues we faced with Devin.”
Hamel Husain, Co-author, Answer.AIAnswer.AI, , write-up
Switched 1
IndeedOn Claude Code
In the talk DX published in June 2026, Indeed's roughly 2,000 engineers worked mostly in Claude Code, after the company retired Copilot and Cody.
“And if you’re interested, our breakdown rate, we’re mostly Claude Code, a little bit Cursor, some Windsurf, and then down at the bottom, Amp. If you actually add up these percentages, it’s definitely more than 100%. So we also see that our devs are using multiple tools at the same time.”
Michael Redding, Principal Product Manager, IndeedDX Engineering Enablement podcast, , episode
Evaluated 1
RunLLMOn Devin
In a month-long trial on the RunLLM codebase, small fixes cost 1 to 5 ACUs, US$2 to US$10, while 2 of 3 larger full-stack tasks gave no usable result (Jan 23, 2025).
“The 3 bigger tasks we tried averaged about 20 ACUs and 2 of the 3 didn’t yield usable results.”
Vikram Sreekanti and Joseph E. Gonzalez, Co-founders, RunLLMFrontier AI, , write-up
Adopted 1
RoxOn Devin
Rox has Devin turn about 350 production conversations a week into evals and open a weekly pull request for the team to review (Oct 14, 2026).
“Every week, Devin, the coding agent from Cognition, reads real conversations on Rox from our production traces in Braintrust, turns them into evals, and opens a pull request that we review.”
Brian Xu, Engineer, Agents team, RoxRox, , write-up
Choosing between Devin and Claude Code
Choose Devin if
Choose Claude Code if
- Engineers want to steer the agent in their own terminal or IDE and approve actions as it works.
- You want the agent inside a Claude Team seat at US$25 per month, or US$20 billed annually, against Devin's US$40 full seat.
- You want one agent across the CLI, IDE extensions, desktop, and web, with project memory in CLAUDE.md.
Side by side
| Dimension | Devin | Claude Code |
|---|---|---|
| Primary mode | Delegated cloud sessions that return pull requests | Interactive sessions in the terminal, IDE, or desktop app, plus cloud sessions on the web |
| Individual priceList prices in US dollars. Included usage differs by plan and is not directly comparable. | Pro: US$20 per month | Claude Pro: US$20 per month billed monthly, or US$17 per month billed annually |
| Team price | Teams: US$40 per full seat per month, US$80 monthly minimum | Claude Team standard seat: US$25 per seat per month billed monthly, or US$20 billed annually |
| Beyond the plan | Prepaid on-demand credits; Enterprise billed in ACUs | Enterprise seats plus usage at API rates; API users pay per token |
| Permissions | Admin security profiles limit a session's network, MCP, and git access; Devin CLI offers an OS-level sandbox | Manual mode starts read-only and asks before edits and commands |
| Team chat | Tag @Devin in Slack or Microsoft Teams | @Claude in Slack on Pro and Max; Claude Tag for shared access on Team and Enterprise |
| Project instructions | AGENTS.md, Skills, and playbooks; Knowledge is deprecated and migrating to Skills | CLAUDE.md, or a repository's AGENTS.md |
| Working together | Accepts handoffs from Claude Code, Codex, and Cursor through the Devin Handoff plugin | Can hand a task to a cloud Devin session with that plugin and a Devin API key |
| Pull requests merged (MSR 2026 study)Public GitHub repositories with at least 100 stars. The authors flag Claude Code's sample of 139 pull requests as small. | 61.6% of 2,252 over 32 weeks from December 2024; 68.0% in the 11 weeks all agents shared; 45.6% of bug fixes19 | 71.9% of 139 over 19 weeks from February 2025; 72.6% of 113 in the shared 11 weeks19 |
What practitioners report, side by side
Each product's figures come from its own practitioners, for the group you pick.
Overall signal
Leaders and teams rate Claude Code +75, 36 points above Devin at +39. Most of the gap is in praise, with 41 of Claude Code's 53 sources positive against 13 of Devin's 24. Claude Code has 1 negative and Devin 4.
The score counts the authors from leaders and teams that take a side, one vote each with their latest view, and is how far positive votes lead negative ones, per 100 votes, so Devin's +39 rests on votes split 13 positive, 6 mixed and 4 negative.
Each of Claude Code's authors wrote one source here, so each source is one vote. A lead of a fifth of the votes reads positive and a lead of half strong positive, with the same lines below zero. The dots show each source, positive, mixed, negative. Use a task pilot to measure accepted work and review time.
Footprint on the Agentic Software Development stack
They differ on three of 12 layers. Claude Code owns two that Devin does not. Coverage is each product's placement on this stack, as its product page shows it. Core is a job the product owns; Supports and Touches are jobs it helps with. Coverage says nothing about how well a product does them.
| Layer | Devin | Claude Code |
|---|---|---|
| Build | ||
| Codebase context | Supports | Core |
| Agent runtime | Supports | Core |
| Validate | ||
| Application security | Supports | Touches |
The 9 layers where they match
| Layer | Devin | Claude Code |
|---|---|---|
| Plan | ||
| Work planning | Supports | Supports |
| Build | ||
| Coding | Core | Core |
| Tool execution | Core | Core |
| Validate | ||
| Debugging | Supports | Supports |
| Testing | Supports | Supports |
| Code review | Supports | Supports |
| Ship | ||
| Source control | Touches | Touches |
| CI/CD | Touches | Touches |
| Operate | ||
| Observability | Touches | Touches |
What practitioners praise
Among leaders and teams, Claude Code alone has four praise themes that reach 5 sources from 3 authors. Large Refactors and Value for Money lead, at 8 of 9 and 5 of 5. Of the two themes both carry, Developer experience differs most. It is praised in 5 of the 7 Devin sources that discuss it, against 9 of 10 for Claude Code. Devin alone has three praise themes that reach 5 sources from 3 authors. Async delegation and End-to-end execution lead, at 9 of 9 and 7 of 8.
| Praised for | Devin | Claude Code |
|---|---|---|
| Codebase understanding | 11 of 11 | 10 of 10 |
| Developer experience | 5 of 7 | 9 of 10 |
| Async delegation | 9 of 9 | Not a theme in its sources |
| End-to-end execution | 7 of 8 | Not a theme in its sources |
| Migrations & repetitive changes | 5 of 5 | Not a theme in its sources |
| Large Refactors | Not a theme in its sources | 8 of 9 |
| Value for Money | Not a theme in its sources | 5 of 5 |
| Multi-Step Execution | Not a theme in its sources | 5 of 6 |
| Enterprise Administration | Not a theme in its sources | 4 of 5 |
Each count is the sources that praise a theme, or raise it as a concern, out of those that discuss it. A row pairs the two products when both carry a theme under the same or an equivalent name.
A count ending in "so far" has under 5 sources from 3 authors among leaders and teams. "Not a theme in its sources" means GILD found no theme of that name in that product's sources.
What practitioners raise as concerns
Among leaders and teams, Claude Code alone has four concern themes that reach 5 sources from 3 authors. Inconsistent Results on Large Tasks and Context Management lead, at 6 of 7 and 6 of 12. Both carry Requires strong review. It is raised in 7 of the 8 Devin sources that discuss it, against 16 of 23 for Claude Code. Unpredictable task success reaches 5 sources from 3 authors only for Devin, at 3 of 6. Claude Code needs a pilot on the same repository tasks before you choose on it.
| Raised as a concern | Devin | Claude Code |
|---|---|---|
| Requires strong review | 7 of 8 | 16 of 23 |
| Unpredictable task success | 3 of 6 | Not a theme in its sources |
| Inconsistent Results on Large Tasks | Not a theme in its sources | 6 of 7 |
| Context Management | Not a theme in its sources | 6 of 12 |
| Usage / Cost | 3 of 4 so far | 5 of 11 |
| Permissions & Safety | Not a theme in its sources | 2 of 10 |
19 sources
- Devin docs: Getting started with Devin Enterprise (SSO)
- Devin docs: Self-serve plans
- Devin docs: Integrations overview
- Devin docs: Linear
- Devin docs: Security profiles
- Devin docs: Devin CLI
- Devin docs: CLI sandbox
- Devin docs: AGENTS.md files
- Devin docs: Skills
- Devin docs: Knowledge (deprecated, migrating to Skills)
- Devin docs: Hand off to Devin Cloud from any agent
- Devin docs: Enterprise deployment
- Claude Code docs: Overview
- Claude Code docs: Platforms
- Claude Code docs: Costs
- Claude Code docs: Security
- Claude Code docs: Memory
- Claude pricing
- Pinna, Gong, Williams, and Sarro, MSR 2026: Comparing AI Coding Agents: A Task-Stratified Analysis of Pull Request Acceptance (Feb 9, 2026)