Interview with Nader Dabit about Devin, Cognition's AI software engineering agent: what makes it different, who it is built for, and why being model-agnostic matters. The discussion also covers coding harnesses, software factories and what is coming next for AI coding.
Cognition's autonomous AI software engineer that runs parallel cloud agents to take tasks from ticket to merge-ready PR.
Updated EU status checked Last shown
Devin at a glance
- What is Devin?
- Devin is an agent from Cognition. Cognition's autonomous AI software engineer that runs parallel cloud agents to take tasks from ticket to merge-ready PR.
- Who makes Devin?
- Devin is made by Cognition.
- Does Devin work in the EU?
- Unverified. No explicit evidence found. Provided pages list enterprise security controls but no supported-countries list, EU data residency, or EU restrictions. Last verified 14 Nov 2026.
- How much does Devin cost?
- Freemium. A pricing page with multiple plans exists, including enterprise options; specific prices and tiers were not captured in the provided text. Licence: Proprietary.
- Where was Devin shown?
- In 2 creator videos by Matthew Berman. Most recently by Matthew Berman on 7 Aug 2026, at 0:22 in “Nader Dabit: Devin, Open-source, Harnesses, Software Factories, and more!”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Devin is an AI software engineering agent from Cognition. It takes delegated engineering tasks, plans work against a mapped codebase, writes code, tests it in its own browser and virtual machines (Windows, Mac and Linux), and prepares pull requests. Larger tasks can be split across parallel cloud agents that Devin manages and reports back on.
It can be used through Devin Cloud, a desktop app, a CLI, or by tagging @Devin in tools such as Slack and Teams. It also integrates with issue trackers and alerting tools to pick up assigned tickets, fix failed CI, triage bugs and investigate alerts through scheduled or trigger-based automations. Devin Review looks for bugs and vulnerabilities in code before production, and Devin Security offers some models, such as Claude Mythos 5.2, to eligible enterprise customers.
Devin is model-agnostic: it combines multiple frontier models, and the vendor claims this gives frontier performance at up to 60% lower cost. Enterprise features listed include RBAC, SSO, SCIM, OIDC, audit logs, per-session VM isolation, optional private networking, and zero data retention enforced with model providers. Specific plan prices were not available in the provided page text.
- Parallel cloud agents that split and manage large tasks
- Own Windows, Mac and Linux VMs with browser testing and session recordings
- Available via cloud, desktop app, CLI, and @Devin in Slack/Teams
- Devin Review for bug and vulnerability detection with one-click fixes
- Scheduled and trigger-based automations (CI failures, alerts, tickets)
- Enterprise controls: RBAC, SSO, SCIM, OIDC, audit logs, private networking
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Recommended as a fantastic option with its own harness and multi-model support.
Similar AI tools
Closest neighbours by meaning, not just by category.
Codex
App · OpenAI
OpenAI's coding agent, available in the ChatGPT app, cloud, IDE extension and CLI under one ChatGPT account.
SWE-2 Flash
Model · Cognition
Cognition's fast, efficient coding model, built for sub-agents handling delegated coding tasks.
Codex in the cloud
App · OpenAI
OpenAI's cloud-based coding agent that runs engineering tasks in remote environments, accessed via ChatGPT.
Codex Environments and Connections
App · OpenAI
OpenAI Codex features for running coding tasks in reusable cloud environments and reaching a desktop Codex remotely.
GPT-5.3 Codex
Model · OpenAI
OpenAI's agentic coding model powering Codex in the CLI, IDE extension, desktop app and ChatGPT cloud.
Codex CLI (refreshed)
App · OpenAI
OpenAI's open-source coding agent that runs in the terminal, with voice control and a new code review experience.