devlo
devlo is an AI software platform and autonomous agent connected to GitHub, Jira and Slack. It builds full-stack apps, edits existing repositories, turns tickets into tested pull requests, reviews code and reports on team productivity for engineering teams and founders.
What is devlo?
devlo is an AI software platform built around an autonomous coding agent that works inside GitHub, Jira and Slack. The publisher positions it as an autonomous AI teammate operating from ticket to deployment, and the product is organised in four blocks. The AI Builder creates full-stack applications from a natural-language brief. The backend is pre-configured, Stripe, email and AI-model integrations can be added in a single instruction, an agent-driven browser runs automated QA, and one click deploys the result to an edge network covering more than 35 regions, with custom domains and automatic SSL. Nine production-ready templates are offered as starting points. The AI Code Editor works on an existing GitHub repository: you chat with the codebase, see a live preview of front-end changes, work on an existing or a new branch, and open a pull request from the conversation, complete with screenshots and a preview link. Integrated QA exercises user journeys and performs visual regression checks. The team agents cover review, tickets and custom automation. The review agent comments on each pull request, offers quick fixes said to take under a minute, supports tag-to-fix from a GitHub review thread, applies review rules defined at organisation and repository level, and reports an applied rate along with the risks it claims to have avoided; devlo advertises a 60% reduction in review time and more than 1,000 engineering teams. The ticket agent analyses the repository, posts an implementation plan in the ticket for human approval, writes the code, runs tests and linters, opens the pull request and answers review comments. Custom agents combine a trigger (new issue, Jira ticket, pull request, failed GitHub check, comment, daily or weekly schedule), written instructions and a set of capabilities. The fourth block is a productivity dashboard: throughput categorised into features, fixes, quality and security, review depth, cycle time broken down by phase, team insights and a quantified view of devlo's own contribution. Frontier models come from Anthropic, OpenAI and Google Gemini, and the choice is left to the user. The publisher reports 70.2% on SWE-bench Verified in May 2025, up from 54.2% in November 2024.
What it does
- Turn a Jira ticket or a GitHub issue into a tested pull request
- Review every pull request automatically and apply the suggested fixes in one click
- Edit an existing GitHub repository from the browser, with a live preview of the changes
- Build a full-stack application from a natural-language brief, on a pre-configured backend
- Build custom agents from a trigger, a set of written instructions and a list of capabilities
- Run automated QA on user journeys, including visual regression checks
- Measure team velocity, cycle time and the impact of AI in a productivity dashboard
When to use devlo / When not to
A quick filter to help you decide if devlo is the right fit.
When to use devlo
- Engineering teams that want to delegate code review, ticket resolution and small fixes to an agent that works directly on their pull requests.
- Founders and small teams that need to go from an idea to a deployed full-stack application without setting up a backend or a build pipeline.
- Engineering managers looking for velocity metrics and a quantified view of what AI actually contributes to their delivery.
- Non-developer members of a product team (product managers, designers, marketers) who need to change something in a repository without waiting for an engineer.
- Teams already standardised on GitHub, Jira and Slack, since every devlo workflow runs through those three integrations.
When not to use devlo
- Teams that do not host their code on GitHub: GitLab, Bitbucket and Azure DevOps are never mentioned as supported forges.
- Anyone who needs a public API to embed devlo into their own tooling: no API or developer documentation is published.
- Developers who require an offline workstation or a desktop application, as the product is exclusively web-based.
- Organisations that require a signed data processing agreement or a stated GDPR compliance position, neither of which is published.
- Users who need an interface in a language other than English, the only interface language available.
How to use devlo
A typical end-to-end flow, from setup to results.
- Create an account from the devlo site by signing in with GitHub.
- Install the devloai GitHub App on the repositories you want the agent to work on.
- Connect Jira and Slack through OAuth from the integrations panel.
- In the devlo app, describe what you want: the agent explores the repository, proposes a plan, writes the code, runs the tests and opens a pull request.
- Alternatively, mention @devloai in a GitHub issue, a Jira ticket or a comment to hand the task over from where you already work.
- On a review thread, comment /fix or tag devlo so it applies the correction and pushes the commit itself.
- To start from scratch, pick one of the nine templates, let the agent generate the application, preview it, then publish it in one click on a devlo.app address or a custom domain.
- Create a custom agent by choosing a trigger, writing the instructions and ticking the capabilities it is allowed to use.
- Teach the agent your conventions in plain language, for example by asking it to remember to run lint checks.
- Monitor consumption in the Organisations panel, where automatic top-up and an optional monthly spending cap can be set.
Pros & Cons
Pros
- Covers the whole delivery chain — building, editing, reviewing, resolving tickets and measuring — where most tools address only one of these.
- Runs from GitHub, Jira and Slack, so the team does not have to change tools to use it.
- The publisher documents its SWE-bench Verified result publicly, reporting 70.2% in May 2025 against 54.2% in November 2024.
- SOC 2 Type II certification, with a public Vanta Trust Center and a published list of subprocessors.
- An opt-out from data training is available from the Pro tier upwards.
- Impact analytics make the return on AI spending measurable, which is uncommon in this segment.
- Free plan and a paid entry point at USD 19 per month, with purchased credits rolling over from month to month.
Cons
- GitHub is the only supported code forge: GitLab, Bitbucket and Azure DevOps are never mentioned.
- No public API and no developer documentation, so devlo cannot be embedded in your own tooling.
- Credit-based pricing makes the real cost of a task hard to forecast: a full-stack app is estimated at 300 to 1,000 credits, that is USD 3 to 10.
- The site never claims GDPR compliance and publishes no data processing agreement.
- All data is hosted in the United States, with no EU region and no European subprocessor.
- The privacy notice acknowledges, in CCPA terms, the sale of data categories through targeting cookies, and the training opt-out is not available on the free or Builder tiers.
- Web-only product in English: no mobile or desktop application, no browser extension and no other interface language.
Pricing & Plans
A permanent free plan is available, and the application also exposes a separate Free (Open Source) tier. The lowest paid entry point is the Builder plan at USD 19 per month, followed by Pro at USD 39 per month and Startup at USD 199 per month; the Enterprise tier is quoted on request. Consumption is metered in credits, with 2,000, 4,500 and 25,000 monthly credits included on the three published paid tiers. Additional credits are purchased at the rate of USD 1 for 100 credits, with a minimum purchase of USD 10; purchased credits do not expire and roll over to the following month, while plan credits are consumed first. Automatic top-up and an optional monthly spending cap can be configured, and payment is made by card through a Stripe checkout. As an order of magnitude, the publisher's FAQ quotes 300 to 1,000 credits for a full-stack app, 300 to 500 for a simple ticket and 100 to 300 for a simple pull request review.
- Free and Free (Open Source) — permanent free plans
- with the Open Source tier listed separately in the application.
- 2
- 000 monthly credits
- full-stack app builder
- support for existing repositories
- live publishing and custom domains
- credit purchasing and raised limits.
- everything in Builder
- 4
- 500 monthly credits
- code review agent
- custom agents
- user management
- all integrations and the opt-out from data training.
- everything in Pro
- 25
- 000 monthly credits
- custom coding agents
- productivity dashboard
- team management and priority support.
- everything in Startup
- VPC deployment
- bespoke integrations and solutions
- onboarding
- enhanced SLA
- dedicated resources and support
- and flexible billing.
- SMB and SMB+ — additional tiers present in the application
- with no publicly listed price.
Data, GDPR & hosting
A consolidated view of how devlo handles your data.
GDPR overview
The site never uses the term GDPR: the acronym appears on no page and in no Trust Center document, so devlo neither claims nor denies compliance. The privacy notice nevertheless addresses EEA and UK rights, mentions the right to lodge a complaint with a local supervisory authority, and links to the list of EU authorities and to the Swiss commissioner. Consent can be withdrawn at any time. On the other hand, no EU representative is designated under Article 27, no data processing agreement is published or offered, all subprocessors are located in the United States, and no transfer mechanism (standard contractual clauses, Data Privacy Framework) is mentioned. Security is documented separately: SOC 2 Type II, with a Type 1 report and a CPA Type 2 report available through the Vanta Trust Center, and published controls covering data classification, retention procedures and deletion of customer data on departure.
Who owns the data?
The privacy notice dated 12 May 2025 states that personal data is collected and processed by devlo.ai. Source code stays with the customer: GitHub is declared as a subprocessor for code storage and versioning, and the homepage FAQ states that "Only the necessary code is accessed and code is deleted upon completion of the task". The subprocessors page adds that "Customers shall retain the unequivocal right to delete any models derived from their data at their discretion, at any time". The Vanta Trust Center lists a "Customer data deleted upon leaving" control, and the account can be closed at any time, with deactivation or deletion of the data held in active databases.
Reuse rights
The privacy notice lists service delivery, authentication, support, feedback, marketing communications, personalised advertising, anti-fraud security and usage analysis as purposes, and describes sharing with advertising networks, performance tracking tools, analytics services, hosting providers and storage and authentication vendors. Under its CCPA section, devlo.ai states that it has sold categories of data through targeting and marketing cookies, and offers a Do Not Sell link. Sensitive personal information is stated not to be processed. To carry out tasks, code is sent to third-party LLM completion APIs (Google Gemini, OpenAI, Anthropic), with zero data retention agreements claimed with those partners. No customer data is used to train devlo's models without explicit consent, and an opt-out from data training is offered as a feature from the Pro tier upwards.
Data retention & training
Hosting summary
Hosting is entirely United States based. The Vanta Trust Center lists Google Cloud Platform as the subprocessor for hosting, security, the database, logging and secrets, located in the USA. The subprocessors page adds Firebase (Google) for authentication, hosting and database, GitHub for code storage and versioning, Google Drive for internal data and Amazon SES for email notifications. Every listed subprocessor is in the United States, and no European region or subprocessor is mentioned anywhere. The devlo web server resolves to 199.36.158.100, an anycast address on AS54113 (Fastly), with the node observed in Mountain View, United States. One distinction matters: the applications that users generate with devlo are deployed on an edge network covering more than 35 regions on the devlo.app domain, which concerns the traffic of those generated applications and not the location of devlo's own data.
Where devlo works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting devlo.
- The site never uses the term GDPR and publishes no data processing agreement, which leaves European buyers without a contractual basis for a data protection assessment.
- The privacy notice acknowledges, in CCPA terms, that categories of data have been sold through targeting and marketing cookies, and offers a Do Not Sell link.
- All subprocessors and all hosting are located in the United States, with no documented transfer mechanism for European data.
- Code is transmitted to third-party LLM APIs (Google, OpenAI, Anthropic) in order to carry out tasks; the zero data retention agreements claimed with those partners are not publicly documented.
- The opt-out from data training is reserved for the Pro tier and above, so free and Builder users do not have it.
- The credit estimates in the FAQ are explicitly given for simple tasks and can be exceeded by a wide margin, which makes budgeting for real work uncertain.
- The corporate name differs between pages — devlo.ai in the terms of service, devlo Inc. in the footer copyright — the SOC 2 reports are only available on request through the Trust Center, and the company is recent (domain registered on 29 February 2024) with no identified funding.
Setup & Integrations
Technical difficulty
Low for a technical team. Nothing is installed locally: everything happens in the browser. Getting started means installing the devloai GitHub App on the chosen repositories, then connecting Jira and Slack over OAuth. Applications created from scratch run on a pre-configured backend, and no model configuration is required, the choice of model being an option rather than a step. The only real friction is having a GitHub account and the right to install an app on the organisation. Advanced settings — review rules per organisation or repository, custom agents, spending caps — are optional and can be added later.
Deployment
Integrations
Supported languages
Behind devlo
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is devlo?
How do I connect devlo to my tools?
Does it work on an existing repository?
Which AI models does devlo use?
How do credits work?
Is there a free plan?
Is my code used to train models?
Where is the data hosted?
Is there a public API or a mobile app?
What results does devlo report on SWE-bench?
Should you pick devlo?
devlo covers an unusually wide stretch of the development chain: building an application from scratch, editing an existing repository, reviewing pull requests, resolving tickets and measuring the result in a productivity dashboard. Most competing tools address one of those tasks; devlo attempts all five, and does so from inside GitHub, Jira and Slack rather than asking the team to adopt yet another workspace. That breadth comes with a single, structural dependency: GitHub. No other code forge is mentioned anywhere on the site, and every workflow, from the ticket agent to the review agent, assumes it. Teams on GitLab, Bitbucket or Azure DevOps are effectively out of scope, and the absence of any public API means devlo cannot be wired into custom tooling either. On the trust side, the publisher does more than most: a SOC 2 Type II certification with a public Trust Center, a published list of subprocessors, a stated position on model training with an opt-out from the Pro tier, and a benchmark result documented in a blog post — 70.2% on SWE-bench Verified in May 2025, presented as the publisher's own measurement. The gaps are equally clear: the term GDPR appears nowhere, no data processing agreement is published, no EU representative is designated, and every subprocessor and hosting location is in the United States. Two other points deserve attention before committing. Credit-based pricing is transparent in its unit rate but hard to forecast for real work, since the published estimates cover simple tasks only. And the publisher is young — the domain was registered in February 2024 — with no identified funding, which matters for an enterprise commitment. For a GitHub-centric team comfortable with US hosting, devlo is a coherent proposition; for a European organisation with formal data protection requirements, the documentation currently falls short.
- Choosing a selection results in a full page refresh.
- Opens in a new window.