Last updated: 2026-08-25
Trylle is a Git platform launched by Trylle, Inc. in 2026 that runs AI agents (Claude, GPT, DeepSeek, and other models) directly inside repositories, executing tasks in a managed sandbox and delivering results as reviewable pull requests instead of routing agent output through a separate third-party app.
About Trylle
Trylle is a Git platform built by Trylle, Inc., a Delaware-incorporated company that launched the product in 2026. It targets engineering teams and open source maintainers who want AI coding agents to work directly inside the same platform that hosts their source code, pull requests, CI, and permissions, instead of bolting an agent on as a separate third-party app with its own credentials. Every agent run in Trylle is triggered by a repository event such as an opened pull request, a failed CI job, or a bot mention, and executes inside a managed sandbox with scoped repository credentials. The agent's work arrives as an ordinary, reviewable, attributable pull request rather than a direct commit, so a human still approves every change before it merges. Trylle lets a team choose which model provider handles a given session (Anthropic Claude, OpenAI, DeepSeek, Moonshot AI, or Z.ai), and its privacy policy states it does not use private repository content to train third-party foundation models unless a customer opts in. Trylle fits engineering teams already living in GitHub who want agent output to land as normal, reviewable pull requests, and open source maintainers who want free security scans on qualifying repositories. Its stacked pull requests, streaming diffs, and a quieter Zen review mode are aimed at teams that ship many small, dependent changes rather than one large branch per feature. The Trylle CLI (try) gives terminal-first developers commands for status, diff, and agent runs with JSON output built for scripting. Trylle offers a free plan with no credit card required, plus two paid tiers priced monthly with larger CI and AI usage grants and more storage; CI runners beyond the included grant are billed per minute of runtime. The platform runs as a web application with a companion CLI for macOS and Linux; native Windows CI runners are not yet available. Trylle launched publicly on Hacker News, positioning itself as an alternative to point solutions like GitHub Actions, Travis CI, and CircleCI; some early readers felt the launch post undersold how the product is different from established platforms such as GitLab, Gitea, or Codeberg. As of this review the company has not published SOC 2, ISO 27001, or GDPR certifications, though its privacy policy commits to not training third-party models on private repository content by default.
Pricing
Free $0/mo (15CI+10AI credit/mo). Pro $4/mo (20CI+15AI). Max $20/mo (30CI+30AI). Linux CI $0.005/min, macOS M4 Pro $0.060/min beyond grants. No annual plans; Windows runners not yet available.
Key Features
- Agent execution inside the platform: Agent sessions trigger on repository events such as pr.opened, ci.failed, or a bot mention, run inside a managed sandbox with scoped repo credentials, and land as a reviewable, attributable pull request instead of a direct commit.
- Multi-model agent sessions: A session can run on any of five supported model providers, including Anthropic Claude Opus 5, OpenAI GPT-5.6 Sol, DeepSeek V4 Pro, Moonshot AI's Kimi K3 Fast, or Z.ai, chosen per task rather than locked to one vendor.
- Two-way GitHub sync: Repositories mirror bidirectionally with GitHub so branches, tags, and commit history stay identical on both platforms, letting a team trial Trylle without migrating away from an existing GitHub repo.
- GitHub-compatible CI: Existing .github/workflows YAML runs unmodified on Linux, ARM64, and macOS runners, with org and repo secrets, build artifacts, and release publishing carried over from GitHub Actions.
- Stacked pull requests: Stacked PRs are a first-class object with a review queue, streaming diffs, and a Zen mode that keeps reviewers inside the branch narrative instead of switching tabs to follow dependent changes.
- Trylle CLI: The try command line tool exposes status, diff, and agent commands with JSON output built for scripting, giving terminal-first developers a daily-driver alternative to the web UI.
Pros
- Trylle's free tier requires no card up front yet still renews CI and AI credit automatically each month, friendlier than trial-based competitors that cut access after a fixed window.
- Because agent work always lands as a normal pull request, a maintainer keeps the same review and approval gate whether the change came from a human or an AI agent using any of the supported model providers.
- Bidirectional GitHub sync lets a team pilot Trylle against a live repository without cutting over history or workflows first, lowering the switching cost compared to platforms that require a one-way import.
Cons
- Hacker News commenters reviewing Trylle's public launch said the landing page did not clearly explain how the product differs from established platforms like GitLab, Gitea, or Codeberg.
- No SOC 2, ISO 27001, or GDPR certification is published on the site, which will matter for regulated or enterprise teams evaluating agent access to their source code.
- Windows CI runners are listed as coming soon, so teams with Windows-specific build or test pipelines cannot run them on Trylle yet.
Frequently Asked Questions
What does Trylle actually cost?
Trylle's Free plan costs $0 a month and includes a $15 CI credit and a $10 AI credit each month, plus 10 GB of storage. Pro is $4 a month with a $20 CI credit and a $15 AI credit and 100 GB of storage, adding CLI AI balance access and AI key export. Max is $20 a month with a $30 CI credit and a $30 AI credit, the same 100 GB storage and extra features. Usage beyond the grants bills per minute: $0.005 for Linux runners and $0.060 for macOS M4 Pro runners, with no yearly commitment on any plan.
Does Trylle have a free plan?
Yes, no card is required to start on the Free plan, which includes unlimited private repositories with 10 GB of storage. Its CI and AI usage grants renew automatically every month but do not roll over, so unused credit is lost at the end of each billing cycle, and once a grant is exhausted that month's CI runs or agent sessions pause until it resets.
What are Trylle's closest competitors?
Cursor and Windsurf compete for teams that want AI pair-programming inside their editor rather than a separate Git platform. Devin competes for teams that want a single autonomous agent to own a ticket end-to-end with less day-to-day oversight. Replit competes for teams that want a hosted environment to build and deploy an app from scratch rather than automate an existing GitHub repo.
What separates Trylle from Devin?
Trylle and Devin approach the job differently: Trylle is the Git platform itself, so any agent's work lands as a pull request inside your normal review and CI queue, while Devin acts as one autonomous engineer working outside your existing platform. Choose Trylle if you want agent output to stay inside your team's existing review process with a choice of model providers; choose Devin if you want a single dedicated agent to carry a ticket further with less oversight day to day.
How do you get started with Trylle?
Sign-up is free and does not require a credit card. Once you connect a GitHub repository, Trylle mirrors it with full commit history through two-way sync, and you can trigger an agent session or import an existing .github/workflows CI pipeline immediately since it is GitHub Actions compatible. The Trylle CLI is optional for teams that prefer a terminal-first workflow over the web UI.
Top Alternatives
- Devin: Pick Trylle if you want agent execution owned by your own Git platform and reviewed as PRs; pick Devin if you want a single autonomous engineer to own a ticket end to end.
- Cursor: Pick Cursor if you want inline AI pair-programming inside your editor; pick Trylle if you want agent-triggered PRs across your whole review and CI workflow.
- Replit: Pick Replit if you want a hosted environment to build and ship an app from zero; pick Trylle if you already have a GitHub repo and want agents working inside your existing CI and review flow.
- Windsurf: Pick Windsurf if you want an AI-native code editor; pick Trylle if you want agent automation triggered by GitHub events without changing your editor.