What each product actually is
Google Antigravity — Google’s agent-first IDE and CLI platform
Google Antigravity is Google’s agent-first development platform. The product thesis is plan-and-execute work inside a Google-native agent desk: Agent Mode, multi-agent orchestration, Gemini-centered and multi-model options, plus the Antigravity CLI for terminal workflows. Catalog places Antigravity in AI-native IDEs and code editors, with a recommended listing and freemium access that escalates into Google AI Pro / Ultra or Google Cloud organization paths.
Antigravity 2.0 (May 2026, per Google’s blog) positions a standalone agent-optimized desktop application alongside the Antigravity IDE path — dynamic subagents, asynchronous tasks, JSON hooks, Scheduled Tasks, project-level grouping, and slash commands such as /goal, /grill-me, /schedule, and /browser. Google’s guidance is to dual-wield the agent-first surface with an IDE of choice (Antigravity IDE or otherwise), not to treat Antigravity as a single “hire a cloud employee” meter.
Buyers are shopping for an interactive Google agent desk — not primarily for a separate autonomous SWE service with its own cloud VM lifecycle. Live tool and review pages: Google Antigravity, Google Antigravity Review. Official: antigravity.google.
Devin — Cognition’s autonomous AI software engineer
Devin is Cognition’s autonomous AI software engineer product line. The thesis is longer-running delegated engineering work: plan, navigate repositories, write and run code, test, open pull requests, review, and collaborate through GitHub, GitLab, Bitbucket, Linear, Jira, Slack, and Microsoft Teams. Surfaces include Devin Cloud sessions, Devin Desktop, CLI handoff into cloud runs, Devin Review, Outposts, and Automations.
Catalog places Devin in AI coding assistants with a recommended listing. Buyers are shopping for Cognition’s autonomous SWE workflow — cloud and desktop agent sessions you supervise and review — not for a Google AI Pro/Ultra IDE seat. Live tool and review pages: Devin, Devin Review. Official: devin.ai.
Same buyer intent, different product types
Both products answer “let AI help build and ship software.” Shared intent is why this comparison exists. Divergent constraints decide the purchase: Google agent-first IDE/CLI desk vs Cognition autonomous SWE platform. Prefer fit over scoreboard geometry.
Current surfaces (per vendor pages)
Surfaces below come from live Cognition and Google pages as of 2026-09-24. They support fit decisions; they are not a substitute for the locked Score table.
Google Antigravity
Agent-first desktop + IDE + CLI. Per Google’s Antigravity 2.0 post, Antigravity ships as a standalone agent-optimized desktop app (macOS, Linux, Windows) with subagents, async task management, JSON hooks, Scheduled Tasks, project grouping, live voice transcription, and slash commands (/goal, /grill-me, /schedule, /browser), while keeping an Antigravity IDE path for developers who dual-wield. The Antigravity CLI (agy) remains part of the platform surface for terminal and automation workflows.
Pricing posture (as of 2026-09-24). Individuals: $0/month with basic weekly rate limits and access to listed Gemini / Claude / open models on the pricing page. Paid individual access is via Google AI Pro / Ultra. Google’s May 2026 plans post describes Pro around $20/month, a $100/month Ultra tier, and a top Ultra tier at $200/month (reduced from $250), with shared Gemini quota pools and AI credits as overage. Organization access is via Google Cloud / Gemini Enterprise Agent Platform consumption.
Sources: antigravity.google · Introducing Google Antigravity 2.0 · Changes to Antigravity Plans · antigravity.google/pricing · Antigravity CLI · changelog
Devin
Devin Desktop (Windsurf successor surface). Cognition’s June 2026 post introduces Devin Desktop as “the next generation of Windsurf”: Agent Command Center as the default IDE surface for managing local and cloud agents, Spaces for shared agent context, ACP support for external agents alongside Devin, and a full IDE retained for last-mile edits. Windsurf is the discontinued predecessor brand — not a third scored member of this comparison.
Fusion harness. Cognition’s September 2026 post makes Fusion available in Devin Desktop and CLI: a lead + sidekick dual-agent harness where you pick a frontier “lead” for planning/review and a cost-effective “sidekick” for execution. Cognition recommends pairing Fable 5.1 with SWE-2 for best results and publishes its own benchmark tables (vendor claims, not aicoolies Scores).
SWE-2. Cognition’s September 2026 SWE-2 announcement positions SWE-2 as Cognition’s most advanced coding model yet, available in Devin Desktop and CLI (with rollout language for Web and Fusion). As of 2026-09-24, the pricing page advertises free use of SWE-2 in Devin Desktop and CLI through 2026-10-10 — a time-bounded promo that expires on that date.
Review / Outposts / Automations. Cognition’s docs release notes document Devin Review expansions (including GitLab and other SCM surfaces), Devin Outposts for running workloads in customer environments (contact sales / enablement language), and Automations with queueing, consumption visibility, and richer triggers. These are product surfaces for buyers evaluating the Cognition line — not Score inputs.
Windows VMs. Cognition documents Windows PC / VM session support (“Devin is Getting a Windows PC”), expanding beyond Linux-only cloud machine assumptions for teams that need Windows toolchains.
Self-serve pricing (as of 2026-09-24). Free $0; Pro $20/month; Max $200/month; Teams $80/month for the team plan + $40/month per full dev seat; Enterprise custom. Included daily/weekly usage quotas with overage at API pricing. Concurrent session limits: Free/Pro up to 10; Max and above unlimited per plan table. Ask Devin, DeepWiki premium modes, and Devin Review usage billing were introduced with the April 2026 self-serve overhaul (OSS Review free language in that post).
Sources: Introducing Devin Desktop · Fusion in Desktop & CLI · Introducing SWE-2 · New self-serve plans · Devin is Getting a Windows PC · docs.devin.ai/release-notes · devin.ai/pricing
Scoreboard (from existing reviews)
Overall: Google Antigravity (score date 2026-09-05) 88; Devin (score date 2026-03-25) 79.
Speed: Google Antigravity (score date 2026-09-05) 91; Devin (score date 2026-03-25) 85.
Privacy: Google Antigravity (score date 2026-09-05) 71; Devin (score date 2026-03-25) 62.
Developer experience: Google Antigravity (score date 2026-09-05) 89; Devin (score date 2026-03-25) 76.
These figures come from the existing aicoolies reviews for each tool. Prefer fit and current product surfaces; use the table as supporting Score references with the dates above.
Antigravity leads overall, speed, privacy, and developer experience on these published totals. That aligns with a polished Google agent-first IDE/CLI story on the 2026-09-05 review. Devin’s 2026-03-25 review remains the published Score reference for Devin — judge today’s Desktop / Fusion / SWE-2 / pricing surfaces on the vendor pages above, and keep the scoreboard secondary.
Neither axis alone settles the buy. A team that needs Cognition’s autonomous cloud/desktop SWE loop with Review and Automations may still pick Devin despite the overall gap. A team that wants Google’s recommended agent-first IDE/CLI path will usually pick Antigravity even when they also evaluate Devin for long-running ticket work.
Legacy note — Windsurf is not Devin’s score
Windsurf is a discontinued listing that rebranded toward Devin Desktop. Its historical Score (score date 2026-03-27, overall 88 on the Windsurf review) belongs to the Windsurf listing and is not Devin’s Score. Those numbers are not Devin’s overall, and they are not parity with Antigravity’s 88. For current Cognition surface area, use the Devin tool page rather than Windsurf.
Workflow, licensing, and privacy
Daily loop
Choose Google Antigravity when the daily loop is an interactive Google agent desk: Agent Mode planning and implementation, desktop agent manager, Antigravity CLI for terminal/automation hooks, and billing through Google AI Pro/Ultra or Cloud org paths. The DX 89 and overall 88 (score date 2026-09-05) support that Google agent-first IDE/CLI story.
Choose Devin when the daily loop is delegated autonomous SWE sessions — Cloud agents, Devin Desktop as the agent command center, CLI handoff, PR Review, Outposts, and Automations — with Cognition’s Free / Pro / Max / Teams / Enterprise packaging. The overall 79 (score date 2026-03-25) is the published review reference; decide on today’s Desktop / Fusion / SWE-2 surfaces first.
Licensing and billing
Antigravity is Google commercial software. Individuals start free with weekly limits; paid tiers buy Google AI Pro/Ultra rate limits and credit overages; orgs go through Google Cloud. As of 2026-09-24, Google’s May 2026 plans post lists Pro around $20/month and Ultra at $100 / $200/month — see also antigravity.google/pricing.
Devin is Cognition commercial software. Self-serve Free / Pro $20 / Max $200 / Teams ($80 + $40/seat) / Enterprise (ACU custom), as of 2026-09-24 on devin.ai/pricing. Usage beyond included quota bills in dollars on self-serve; Enterprise continues ACU language per Cognition. The SWE-2 free-through-2026-10-10 Desktop/CLI promo is time-bounded and expires on that date.
Privacy posture on these Scores
On the published Score table, Antigravity leads privacy (71 vs 62). Both privacy Scores sit softer than many OSS/BYOK CLI agents on aicoolies — expected for commercial hosted Google and Cognition paths. Read these as relative strengths: Antigravity’s privacy 71 is the stronger of the two on these reviews; neither product is a local-only OSS agent story like Goose or Aider.
Pricing and packaging — buyer choice forks
Antigravity packaging. You buy Google AI Pro/Ultra (or Cloud org consumption) to raise rate limits and concurrency on an agent-first Google desk. The free tier exists but weekly limits matter. Price is not “ACUs for a cloud employee”; it is Google subscription / Cloud consumption for agent sessions inside Antigravity’s surfaces.
Devin packaging. You buy Free / Pro / Max / Teams / Enterprise for Cognition’s autonomous SWE line, including Desktop, Cloud, CLI, Review, and automations. Teams adds collaboration and centralized billing; Enterprise adds SSO, dedicated deployment options, and ACU billing. Concurrent session limits and quota refresh cadence matter as much as list price for heavy agent users.
Practical fork. If procurement already standardizes on Google AI / Gemini Enterprise, Antigravity is the natural desk. If procurement already evaluates Cognition for delegated PR/ticket workflows — or needs Desktop as the Windsurf successor IDE plus Cloud agents — Devin is the natural line. Each product has its own packaging story; forcing one into the other’s billing model usually misfits the workflow.
Who should pick which
Pick Google Antigravity if you want:
Google’s agent-first IDE/CLI platform (Agent Mode, desktop agent manager, agy CLI)
Gemini-centered and multi-model options on Google AI Pro/Ultra or Cloud org billing
A recommended Google-path listing for interactive agent desks
The stronger overall / speed / DX Score profile on these reviews (88 / 91 / 89, score date 2026-09-05)
Pick Devin if you want:
Cognition’s autonomous SWE product line — Cloud sessions, Devin Desktop, CLI handoff, Review, Outposts, Automations
Longer delegated engineering sessions with PR-centric collaboration
Self-serve Free / Pro $20 / Max $200 / Teams packaging (or Enterprise ACUs)
Fusion lead+sidekick harness and SWE-2 availability as documented by Cognition — judged on current surfaces, with Score dated 2026-03-25 (overall 79) as the review reference
Pick neither as a Windsurf Score substitute. Windsurf’s historical 88 belongs to the Windsurf listing and is not Devin’s Score.
Editorial winner: google-antigravity — fit-first Google agent-first IDE/CLI default, with a clear overall Score lead (88 vs 79) that aligns with that pick but does not replace product-shape judgment. State Devin’s autonomous SWE niches explicitly so the page does not read like a fake landslide.
How this fits the Antigravity / Devin neighborhood
On aicoolies, Antigravity already sits next to Cursor, Kiro, Trae, and Zed comparisons in the AI-native IDE / editor shelf. Devin sits next to Claude Code, Cursor, and OpenHands in the autonomous / agent coding shelf. This page answers the direct buyer question after Windsurf’s discontinuation: Google’s agent-first platform versus Cognition’s live Devin line.
Related live peers include google-antigravity-vs-cursor, google-antigravity-vs-kiro, google-antigravity-vs-trae, google-antigravity-vs-zed, and Devin vs Claude Code / Cursor / OpenHands. Use those for other forks. Use this page when the fork is specifically Google agent-first IDE/CLI vs Cognition autonomous SWE.
Windsurf is not part of the winner or scoreboard on this page. It appears only as discontinued predecessor context for Devin Desktop.
Decision guide — five practical forks
1. Do you need a Google agent-first IDE/CLI desk? If your team already buys Google AI Pro/Ultra or Gemini Enterprise Agent Platform, Antigravity is the natural interactive agent surface. Devin can still sit beside it for delegated cloud tickets, but that is coexistence — not a substitute for Antigravity’s Google desk.
2. Do you need Cognition’s autonomous SWE loop with Review and Automations? If the job is long-running delegated PRs, cloud VMs, Desktop agent command center, Outposts, and automation triggers, Devin is the product line. Antigravity’s Agent Mode is interactive Google agent work — different shape.
3. Is “Windsurf” still in the RFP language? Translate Windsurf → Devin Desktop. Windsurf’s historical Score belongs to the Windsurf listing and is not Devin’s Score. Evaluate Devin on Desktop / Cloud / CLI / Review / pricing pages.
4. Do you optimize for interactive DX on a Google path? Antigravity’s DX 89 and overall 88 (score date 2026-09-05) support the “desk you live in” Google agent story. Devin’s DX 76 (score date 2026-03-25) sits behind that on the published table — fine for teams that prioritize autonomous session throughput over Google IDE feel.
5. Are Scores enough to decide alone? No. Product shape and packaging decide. Use the locked Score table with both dates as secondary review evidence. Prefer Antigravity when you want the Google agent-first default; prefer Devin when you want Cognition’s autonomous SWE line.
What “good” looks like for each buyer
Antigravity buyer profile
A strong Antigravity buyer already trusts Gemini-class models and Google Cloud packaging. They want an agent-first desk — plan, implement, review artifacts, run scheduled agents — with CLI escape hatches. They are willing to pay Google AI Pro/Ultra (or Cloud org) for rate limits and concurrency. They compare Antigravity to Cursor, Zed, Kiro, and Trae more than to cloud-employee SWE services.
Devin buyer profile
A strong Devin buyer wants to delegate well-scoped engineering work to Cognition agents and review the PR. They care about Cloud + Desktop + CLI continuity, Review quality, Automations, and team seats. They compare Devin to OpenHands, Claude Code (for different reasons), and other autonomous assistants — and they should evaluate Desktop / Fusion / SWE-2 on today’s vendor pages rather than treating March Score alone as the full product picture.
Shared intent, divergent constraints
Both answer “AI helps ship software.” Shared intent is why the comparison exists. Divergent constraints — Google agent desk vs Cognition autonomous SWE — are why the winner is fit-first rather than “always pick the higher overall Score.”
Feature surfaces that matter in practice
Agent orchestration
Antigravity emphasizes interactive multi-agent orchestration inside Google’s desk (subagents, scheduled tasks, slash-command control). Devin emphasizes delegated session autonomy with lead/sidekick Fusion options and multi-session Cloud/Desktop management. For the latest slash-command and Fusion UI labels, see the Google and Cognition pages linked in Sources.
Editor vs cloud employee
Antigravity keeps you in a Google agent/IDE loop. Devin keeps you in a Cognition session/PR loop, with Desktop providing an IDE shell when you need last-mile edits. Buyers who want “stay in my Google desk” pick Antigravity. Buyers who want “delegate the ticket and review the PR” pick Devin.
Review and automations
Devin’s Review, Outposts, and Automations are first-class Cognition surfaces for team workflows. Antigravity’s strength is agent-first Google development with scheduled tasks and hooks — different automation shape. Feature lists belong on the vendor pages linked below rather than as invented parity tables.
Model posture
Antigravity is Gemini-centered with documented multi-model options on free/paid tiers. Devin is model-flexible inside Cognition’s product (including SWE-2 and Fusion pairings per Cognition). Neither Score table is a model-quality league table.
Cost and procurement notes
Antigravity cost shape. As of 2026-09-24, Google’s May 2026 plans post describes Google AI Pro around $20/month and Ultra around $100 / $200/month, with a free individual tier that has weekly limits, or Cloud org consumption. See Changes to Antigravity Plans and antigravity.google/pricing.
Devin cost shape. As of 2026-09-24, expect Free / Pro $20 / Max $200 / Teams $80+$40/seat / Enterprise ACU on devin.ai/pricing. Quota + overage predictability matters; Ask / Review / DeepWiki usage billing since the April 2026 self-serve change. The SWE-2 free Desktop/CLI promo runs through 2026-10-10 and then expires.
Total cost of ownership is not overall Score. A team can spend more on Max or Ultra and still be correct if the product shape matches the workflow. Locked Scores help compare reviewed fitness; they are not an invoice calculator.
Privacy and data-path framing
On published aicoolies Scores, Antigravity leads privacy (71 vs Devin 62). Read that as review evidence for relative posture on commercial hosted paths — not a certificate that either product is local-only or air-gapped by default.
Practical privacy questions for buyers:
Where do prompts and repository context go by default?
Can you keep work inside Google Cloud org controls (Antigravity) or Cognition Enterprise / Outposts controls (Devin)?
Does your compliance team accept Google AI / Cloud terms or Cognition commercial terms?
Do you need a local OSS agent instead? If yes, this page is the wrong fork — look at Goose, OpenCode, Aider, Trae Agent, etc.
Antigravity answers (1)–(2) inside Google’s commercial path and scores 71 privacy (score date 2026-09-05). Devin answers (1)–(2) inside Cognition’s commercial path and scores 62 privacy (score date 2026-03-25).
Speed and DX — what the leads mean
Antigravity leads speed (91 vs 85) and DX (89 vs 76) on the locked Score table. In buyer language: the Google agent-first desk felt faster and fit the interactive developer loop better in those reviews.
Devin’s lower speed / DX Scores on the 2026-03-25 review are not a claim that Devin is “bad.” They mark a different optimization. Devin was reviewed on 2026-03-25, before those launches (Desktop, Fusion, SWE-2). If your evaluation rubric weights autonomous PR throughput and Cognition team surfaces above Google IDE feel, Devin can still win the purchase even while trailing those Score axes.
Overall 88 vs 79 is a clear gap on paper. Fit-first still applies: an org that only wants Cognition’s autonomous SWE line should not auto-buy Antigravity. Prefer Antigravity when the org wants Google’s recommended agent-first IDE/CLI path.
Migration and coexistence patterns
From Windsurf to Devin. Windsurf users should evaluate Devin Desktop as the successor surface and the Devin tool page as the live product page. Windsurf’s historical Score belongs to the Windsurf listing and is not Devin’s Score.
From Antigravity interest to Devin. Teams often keep Antigravity for interactive Google agent work and add Devin for delegated cloud tickets. That coexistence is coherent.
From Devin interest to Antigravity. Teams standardized on Google AI / Cloud may prefer Antigravity as the daily agent desk and use Devin only if Cognition-specific Review/Automations are required.
Editorial winner on this page remains google-antigravity. Coexistence is a team topology choice, not a winner flip.
FAQ-style clarifications
Is Windsurf the same as Devin? No. Windsurf is a discontinued predecessor listing. Devin Desktop is the successor surface. Scores are not transferable — Windsurf’s historical Score belongs to the Windsurf listing and is not Devin’s Score.
Does the overall Score mean Antigravity always wins? No. Antigravity is the fit-first winner and leads overall 88 vs 79, but Devin remains the better pick for Cognition autonomous SWE buyers.
Can I use Devin Desktop like an IDE? Cognition positions Devin Desktop as retaining a full IDE while making the Agent Command Center the default surface, as of 2026-09-24 on the Desktop announcement and product docs.
Are Free / Pro / Max / Teams still current? Yes on devin.ai/pricing as of 2026-09-24 (Free $0, Pro $20, Max $200, Teams $80+$40/seat).
What about Google AI Pro / Ultra dollars? As of 2026-09-24, Google’s May 2026 plans post describes Pro around $20/month and Ultra at $100 / $200/month. See Changes to Antigravity Plans and antigravity.google/pricing.
