Spec-to-PR autopilot
Drop a markdown spec in .gluecron/specs/. Claude reads it, opens a draft PR with the implementation, requests review — all within 90 seconds. Zero clicks between idea and diff.
See it runAhead: server-rendered speed, honest status, AI you can switch off, post-merge verification. Level: git, PRs, issues, packages, pages. Behind: CI ecosystem and scale, security feeds, enterprise certifications. Stated plainly — a preview, not a launched product.
Every row cross-referenced against the public parity scorecard. When GitHub legitimately wins on a row, we say so — honesty makes the wins land harder.
Not "soon". Not "in beta". Not "via marketplace". Just no. Each one of these is live on Gluecron right now.
Drop a markdown spec in .gluecron/specs/. Claude reads it, opens a draft PR with the implementation, requests review — all within 90 seconds. Zero clicks between idea and diff.
See it runTalk into your phone. MediaRecorder ships the audio to Claude, Claude opens a PR. Diff is live in seconds — while you're still holding the phone.
Try voice modeOne brief in English. N PRs across N repos. "Rename trackEvent to logEvent everywhere" → eight PRs queued, eight reviewers tagged.
Open the consoleFigma-style cursors on the diff. See your teammate's selection in real-time. Resolve threads together without the merge-conflict dance.
Watch the demoPer-agent identity with agt_… tokens. Scoped leases prevent two agents from clobbering each other. The first git host built for AI teammates.
Issue agent tokensConfidence-scored heuristics hold every new commenter for review before their first comment lands; maintainers approve or block from a queue. No AI required.
See the shieldTo match Gluecron Pro feature-for-feature, you'd be running six separate subscriptions. Here's the receipt.
GitHub gives one PAT per human. Gluecron issues per-agent tokens with scoped leases, so a swarm of agents can work the same repo without stepping on each other.
# 1. Issue a per-agent token (NOT a per-user PAT) $ gluecron agents create --name claude-nightly --scope repo:write agt_2k9Lpqr...4xZ # 2. Lease a file before editing — no two agents touch it twice $ gluecron lease acquire src/api.ts --ttl 5m lease_a1b2... acquired (expires 18:42 UTC) # 3. Push as the agent — appears as its own author in history $ git -c user.name="claude-nightly" push gluecron main remote: ✓ lease released ✓ gates green ✓ deploy queued # 4. Watch the swarm work in real time $ gluecron agents tail [claude-nightly] +47 / -12 src/api.ts merged #421 [claude-deps] bumped 6 deps merged #422 [claude-issues] built issue #88 → PR opened #423
GitHub Enterprise Server is a virtual-appliance image that takes professional services to install. Gluecron self-host is one curl.
curl -fsSL https://gluecron.com/install-server | bashOperations playbookFair point — GitHub's Actions marketplace is years ahead. Gluecron's workflow runner uses the same yaml shape and runs on your server, so most workflows port directly. For the long-tail of third-party actions, you're on your own for now; that's the honest trade.
Claude's PR review posts inline comments — you're free to ignore them. Auto-merge is opt-in per branch protection rule and re-uses every gate your manual merge already enforces. If Claude blocks a good PR, override is one click; if Claude approves a bad one, your required checks still have to pass.
Yes. `/import` clones a single repo via PAT; `/import-bulk` mirrors an entire org in one pass. You can keep pushing to GitHub during the cutover and flip the default remote when you're ready. A migration verifier checks object counts and branch parity post-clone.
Because it's open, single-tenant, and self-hostable. Your code lives on your disk in your database. If we vanish tomorrow, your git history is already on your machine — and the gluecron binary will keep working without our servers. That's not a promise GitHub can make.
GitHub: not possible. Gluecron: built-in. Auto-merges fire the instant gates go green, ai:build issues become PRs in 90 seconds, and failing gates get patched in real time. Toggle Sleep Mode if you'd rather batch it and catch up async — but the speed is always there.
One weekend. Twenty-five-plus capabilities. Nineteen dollars. The honest scorecard makes the choice obvious.