Talyn

Talyn

Talyn runs Claude Code and Codex in parallel cloud microVMs on your own Claude or ChatGPT subscription, one branch per task, then opens the pull request and fixes failing CI, conflicts and review comments until it merges. Free with 10 box-hours a month; Pro is $40.

0.0 (0)
1 Impressions
Visit Website
Founder
Tom Owers (Talyn)
Active
PlanFree - $0 - 10 box-hours/month, small and standard boxes, 2 running at once · Pro - $40/mo or $400/yr - 200 box-hours, all box sizes, up to 100 boxes, auto-keep and auto-review, credits; Team $40 per seat · Max - $100/mo (monthly only) - 1,000 box-hours, all Pro features
BusinessB2B

About Talyn

What Talyn is

Talyn runs Claude Code and Codex in parallel cloud sandboxes and then carries each change all the way to a merged pull request. Every task becomes a session with its own branch, agent, Firecracker microVM, diff and terminal, so you can start as many as you have tasks, close your laptop, and come back to a diff or to a pull request that is already green. The agents run on the Claude Pro/Max or ChatGPT Plus/Pro subscription you already pay for: Talyn charges for the machines (box-hours), never for tokens. It works with GitHub only, through a GitHub App, and comes as desktop apps for macOS, Windows and Linux plus a web app at app.talyn.dev.

Talyn is built by one developer, Tom Owers, in public: the source is MIT-licensed at Gilbert09/talyn on GitHub, and the about page says plainly that it is not a company with a team behind it. It launched on Product Hunt on November 15, 2026 ("Run Claude Code and Codex in parallel cloud sandboxes") and finished fourth for the day. The timing is not accidental: Conductor Cloud shuts down on November 20, 2026, and Talyn reads conductor.json and .conductor/settings.toml unchanged.

How it works

  • Sessions. One task on one branch in its own sandbox: chat, diff, files, terminal and run scripts. Sessions are isolated from each other, so two agents never edit the same files. In the desktop app you can also run a session in a local git worktree, or move a Claude Code conversation from your computer into a sandbox and carry on.
  • Threads. A conversation with no repository of its own. Give it a goal ("show usage in the app") and it starts a session on every repository that needs a change, then follows each one; a session that needs a decision asks you in the thread.
  • Pull requests to green. Talyn opens the PR, then fixes failing checks, merge conflicts and review comments on the same branch. Flag a PR "keep mergeable" and an agent fixes it whenever it falls behind, conflicts or goes red; the merge queue lands it in order when it is green. Merges go through GitHub's normal merge API, so branch protection and required reviews still apply.
  • Automations. GitHub automations run on pull request events (label, request a reviewer, comment, send an agent, queue it); scheduled automations run a prompt on a repository at a set time; Slack reactions, mentions or DMs start a session with the thread as the brief. Linear, GitHub Issues and webhooks are also supported, and connected MCP servers (Linear, Sentry, Supabase or your own) are available to every run.
  • Code review. Several agent reviewers read a change from different angles, a checking pass throws out findings it cannot stand behind, and the short list stays in the app until you choose "Post to PR". Your own SKILL.md playbooks (from the repo, ~/.claude/skills or the workspace) can run against any PR.
  • PR dashboard. One live list of every open PR across connected repositories, blocked ones first, plus review requests.

Security model

The sandbox never holds your agent sign-in or MCP keys. When the agent calls the Claude or OpenAI API, or an MCP server, the request goes through Talyn's proxy outside the sandbox, which attaches the credential, so an agent tricked by a hostile file has no keys to leak. Stored tokens are encrypted with AES-256-GCM; API tokens are stored only as hashes. Sessions have internet on by default; with it off, a sandbox reaches only its repository, the package registries and the agent's API, and fix runs, reviews and scheduled automations are restricted that way by default. A sandbox cannot reach your computer or a private network, and Talyn deletes it when you archive the session. Unpushed work is saved to Talyn's storage after every turn so a lost sandbox can be restored.

Pricing (checked November 16, 2026)

PlanPriceBox-hours / monthLimits
Free$010Small and standard boxes, 2 running at once, keep-mergeable on PRs you flag
Pro$40/mo or $400/yr200 (about $0.20 each)All box sizes, up to 100 at once, auto-keep and auto-review on every new PR, credits
Max$100/mo (monthly only)1,000 (about $0.10 each)As Pro
Team$40 per seat200 per personOne invoice; no shared workspaces, SSO or admin yet

A box-hour is one standard box (4 vCPU, 4 GiB) awake for an hour, billed per second. Small boxes (2 vCPU, 4 GiB) count half; medium (4 vCPU, 8 GiB) 1.25; large (6 vCPU, 12 GiB) 2; XLarge (8 vCPU, 16 GiB) 2.5. Above the allowance, credits pay $0.22 per standard box-hour, sold in $25, $100 (+5%) and $500 (+10%) packs that last 12 months. Sleeping boxes are free. Your Claude or ChatGPT subscription is billed separately by Anthropic or OpenAI.

Worked example: which plan does a real workload need?

Using the published rates, we sized a typical solo workload: 8 agent sessions per working day, each awake about 20 minutes on a standard box, over 22 working days, plus a nightly scheduled automation that runs about 10 minutes on a small box (automations default to small).

sessions:     8 x 20 min x 22 days = 3,520 min = 58.7 h x 1.0  = 58.7 box-hours
automation:   10 min x 30 nights   =   300 min =  5.0 h x 0.5  =  2.5 box-hours
total                                                           = 61.2 box-hours/month

Free  (10 included)  -> 51.2 over; credits are Pro/Max only -> does not fit
Pro   (200 included) -> fits, ~139 hours spare            -> $40/month
on credits alone     -> 61.2 x $0.22                      ~= $13.46

The takeaway: the free plan is a trial for a few sessions a week, not daily use, and because credits are only sold on paid plans, Pro is the realistic entry point for daily work. Memory matters too: a box is billed at the Standard rate while it uses 4 GB or less, so a Large box only costs the Large rate while it actually uses more than 8 GB.

We checked the repository

GET https://api.github.com/repos/Gilbert09/talyn   (2026-11-16)
created_at: 2026-04-16    stars: 19    forks: 5    open issues: 3
license (API): NOASSERTION   # LICENSE file: "The MIT License (MIT), Copyright (c) 2026 Tom Owers"
languages:  TypeScript (~28 MB), Python, JavaScript, PLpgSQL, CSS, Shell, Dockerfile
releases:   v0.3.94 2026-11-14, v0.3.95 2026-11-15, v0.3.96 2026-11-16

GitHub's API reports the licence as NOASSERTION only because the LICENSE file adds a note that parts of the desktop scaffolding come from Electron React Boilerplate (also MIT); the file itself is the MIT licence. Releases ship roughly daily, which matches the "public beta" label.

Talyn vs the alternatives

TalynCodex CLI / Claude Code on your laptopLetter
Where the agent runsCloud microVM per sessionYour machineVendor's coding agent
Model billingYour Claude/ChatGPT plan; Talyn bills machinesYour plan or API keyToken plans
Ends atMerged PR (fix CI, conflicts, merge queue)A local diffOne PR per edit of your live site
ParallelismUp to 100 boxes on ProLimited by your laptopn/a

For a head-to-head of the underlying agents, see our Claude Code vs Codex CLI vs Gemini CLI comparison. Talyn's own site compares it with CodeRabbit, Codex cloud, Cursor cloud agents, Devin, Graphite, Greptile, Mergify and T3 Code; those pages are vendor-written.

Who it's for

Developers and small teams already paying for Claude Code or Codex who want several agents working at once without keeping a laptop awake, people drowning in red or stale pull requests, and Conductor Cloud users who need a home before November 20, 2026.

Limits and trade-offs

  • GitHub only, with no near-term plan for GitLab or Bitbucket.
  • No team features yet. Billing is per account; shared workspaces, SSO and admin controls don't exist, and the maker says it probably won't clear a company tool-approval process today.
  • One-person public beta. Support is through GitHub issues; Windows installers are unsigned, so SmartScreen warns on first install.
  • Write access to code. The GitHub App needs write access so agents can push fixes. Auto-merge is opt-in per PR, but review what you flag.
  • Small inconsistency in the docs: the pricing table says a box sleeps after 10 idle minutes, while the FAQ and sandbox docs say 3 minutes.
  • Two bills. Talyn's box-hours plus your Anthropic or OpenAI subscription, whose own usage limits still apply to every session.

Related on TechLogHub

Value & Audience

Value Proposition

Run many coding-agent sessions at once without keeping your laptop open, pay only for sandbox time on the AI plan you already have, and get each change to a green, merged pull request.

Problem Solved

Coding agents tie up the developer's laptop one task at a time, and the pull requests they open go stale or red on failing checks, conflicts and review comments that still need a human to babysit.

Share this product

Product founder?

Get Badge

Tech Stack

Social & Links

Loading ratings...

Loading comments...

Gallery

Talyn screenshot 1

Feedback & Roadmap

Loading feedback…

Talyn FAQ

Common questions about Talyn's features, pricing, and use cases

Is Talyn free to use?

Talyn has a free tier plus paid plans, starting at Free - $0 - 10 box-hours/month, small and standard boxes, 2 running at once. Plans listed: Free - $0 - 10 box-hours/month, small and standard boxes, 2 running at once; Pro - $40/mo or $400/yr - 200 box-hours, all box sizes, up to 100 boxes, auto-keep and auto-review, credits; Team $40 per seat; Max - $100/mo (monthly only) - 1,000 box-hours, all Pro features.

What are some alternatives to Talyn?

Popular Developer Tools alternatives to Talyn include CodeSpy, DevOp and Global TTFB Checker.

Talyn appears in

Newsletter

One email a week

TechLogHub tracks 874 tools and 599 open-source projects. Get the week's new launches, release roundups and open-source picks — nothing else.

Get the TechLogHub digest

One email a week: new and trending developer tools, fresh comparisons, and what shipped. Unsubscribe in one click.