The Git platform where agents ship like engineers.

Trylle is an AI-powered Git platform where agents handle PR reviews, security audits, and repo automations.

Get started - it’s free!See pricing
#8
6+24-0
Pull request #8

Add support for hidden flags in shell completion

This PR adds completion generation for the nimbus command across bash and zsh shells, with opt-in support for hidden flags via NIMBUS_COMPLETE_HIDDEN=1 environment variable.

The implementation introduces a new completion.go module that gates hidden flag visibility during completion generation. A seed GenBash() function in completion_gen.go provides bash completion output, while _nimbus_bash_complete completion is disabled by default for hidden flags; export NIMBUS_COMPLETE_HIDDEN=1 to reveal them.

The change is additive and backward-compatible — existing completion behavior stays intact. Test coverage validates that the TestHiddenFlagsMaskedByDefault function is callable.

01 / 06

Add zsh completion stub for nimbus

+2 -0

contrib/completion/_nimbus.zsh

Registers a minimal _nimbus completion function so shells can discover subcommands without sourcing the full CLI.

contrib/completion/_nimbus.zsh+2
02 / 06

Teach bash about hidden completion flags

+4 -0

contrib/completion/_nimbus.bash

Bash completion now forwards NIMBUS_COMPLETE_HIDDEN when set, mirroring zsh behavior for scripting environments.

contrib/completion/_nimbus.bash+4
03 / 06

Generate completion metadata in build

+6 -0

cmd/nimbus/completion_gen.go

Embeds a static table of commands and hidden flags into the binary so offline completion stays in sync with nimbus --help.

cmd/nimbus/completion_gen.go+6
04 / 06

Document hidden flag env vars

+3 -0

docs/cli/completion.md

Operators can opt into exposing experimental flags during completion by exporting the documented variables.

docs/cli/completion.md+3
05 / 06

Wire completion tests

+5 -0

tests/completion/hidden_test.go

Golden tests assert that hidden flags stay out of default completion output unless the env gate is on.

tests/completion/hidden_test.go+5
06 / 06

Changelog entry

+4 -0

CHANGELOG.md

Release notes for completion improvements and the new env toggles.

CHANGELOG.md+4

Simple, affordable platform pricing

Pay as you go using published rates. No sketchy billing structures, no yearly commitments, no hidden fees.

per minute
Linux x64
Linux ARM64
macOS
Windows
Free CI /mo
Free AI /mo
Choose model
No yearly commitment
$0.005
$0.005
$0.060(M4 Pro)
Soon
≈3,000 min
$10
$0.012
$0.008
$0.062(M1)
$0.022
2,000 min
$0.020
$0.020
$0.060(M1)
400 min

GitHub and GitLab rates for their 4 vCPU hosted runners, from their public billing documentation, August 2026.

Estimate your cost of shipping

Move the sliders. Everything inside your plan's included credit costs $0. Anything on top is pay as you go at published rates.

Plan
$0/mo
CI minutes3,000 minutes/mo

3,000 minutes included — $0

About 50 hours of 4 vCPU computing

AI tokens133M tokens/mo

133M included — $0

About 10 hours of thinking

CI uses the Linux 4 vCPU rate of $0.005/min. Cost per 1M tokens is estimated based on average usage of DeepSeek V4 Flash Fast (ZDR); actual model, settings and token caching determines the real cost. Hours of thinking metric is estimated from average tokens cost per 1 hour of inference

Adopt it without moving off GitHub

Two-way sync keeps branches, tags, and history identical on both remotes. Teammates who never create an account keep pushing to GitHub, and their commits show up in your review here.

Sync runs both directions

Push to either remote. Both stay current, so nobody is cut off mid-sprint.

Import, don't re-create

Public or private repositories come across with their full history intact.

Discover why our GitHub Sync is awesome

Run models where the code already is

A prompt, a trigger, a model, and a permission scope. Automations run inside the platform on real repository events, using scoped repository credentials — no third-party app to install, no token handed to someone else's cloud.

Any frontier model, one balance

Frontier models from every major provider — pick one per automation, with dedicated ZDR routes when code cannot touch a retaining provider. Hosted model usage draws from your AI credit and wallet balance.

Triggered by repository events

A pull opens, CI fails, a bot is mentioned — the run and its output land on the timeline. Six official templates install in one click; edit the prompt or start from an empty one.

Zero-data-retention routing runs through Vercel AI Gateway and Wafer.

A complete delivery platform, not a wrapper

Everything you expect from a software-delivery foundation ships here — in the same system your agents run in.

  • Git repositories

    Browsing, commits, compare, and archives.

  • Pull requests

    Review threads, streaming diffs, Guided PRs, and drafts.

  • Issues

    Labels, mentions, timeline, and notifications.

  • Actions

    GitHub-compatible CI runs your existing .github/workflows, with secrets, variables, and artifacts.

  • Stacked pull requests

    Parent branches are first-class, so stacks build and rebuild without a second tool.

  • Releases

    Tags, notes, and assets.

  • Orgs & teams

    Members, roles, repository roles, org-level bots and automations.

  • The try CLI

    Repositories, CI, review, stacks, and agent handoff from the terminal.

  • Two-way GitHub Sync

    Branches, tags, and history stay identical on both remotes.

On the roadmap:Integrations MarketplaceWindows runnersModerationOrganization webhooksMore Integrations

Trylle CLI is your daily driver

Each command starts from the repository you are already in. See what needs attention, act on it, and return to the code without rebuilding context in another tab.

Start with the work queue

try status combines your identity, current repository, unread inbox, review requests, open pull requests, recent issues, CI, automations, and invitations.

Review at terminal speed

try diff opens a side-by-side TUI for working changes, commits, ranges, stacks, or pull requests—with file progress, themes, watch mode, and inline annotations.

Close the agent loop

Annotate a selection, hunk, or file, then press s. The CLI sends clean review notes through stdout so Codex, Claude, or any shell-capable agent can take the next turn.

Discover why our Trylle CLI is awesome

Start with $25 of free grant every month

Free accounts get a usage grant every month, not a trial. It renews on the first, needs no card, and spends on whichever of CI or AI you actually use.

$15 for CI every month

About 3,000 free minutes of Linux, 4 vCPU, or the same balance spent on ARM64 or macOS if that is what your matrix needs.

$10 for AI every month

Every model in the catalog draws from it, including models on dedicated ZDR routes. After AI credit is used, usage draws from your pay-as-you-go balance.

We ♥ Open Source Community

Helping OSS projects ship safely without drowning maintainers in agent spam

Safe distribution

Trust follows every release. Reviewable human and agent changes, repository-owned checks before merge, and releases traceable back to source.

Spam-safe collaboration

Maintainers stay in control. Trusted identities for people and agents, automated issue and pull request triage, and maintainer-controlled agent access.

Free security scans

Trylle sponsors qualifying open-source projects with free agentic audits for security issues and other bugs. Reach out for a free scan today.

Contact us

Something you can try today

Agent execution, review, permissions, and CI in one system. No credit card and no trial time. Just free.

Get started - it’s free!See pricing