GoLive Launch Assistant
Let your coding agent plan and deploy an app, with your approval before it changes real cloud accounts.
- Source repo
- mikehasa/golive-skill
- Stars
- ★ 1.6k
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 81/100 · Good
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- High · needs real infrastructure
- You'll need
- Typical use
- A developer using Codex or Claude Code who wants to deploy an existing app to Vercel or Netlify.
- Not a fit if
- Users who need a validated first-account setup
- Teams requiring live validation across all app frameworks and provider pairings
- Website
- trytofu.ai ↗ set on GitHub
- Maintenance
- Commits in the last 6 monthslast commit 2026-10-11latest release v0.1.0-alpha.9 (2026-10-11)+5.9 ★/day over 7 days
- Source review
- 81/100 · Good
What does this agent do, and when should you use it?
GoLive is an open-source Agent Skill with a zero-dependency Node.js CLI for assessing and planning an app launch from its repository. The conversational skill guides the workflow, while the CLI handles detection, planning, approved changes and verification. Built-in paths cover Vercel or Netlify hosting, Supabase or Neon databases, and live-tested DNS, Resend email, Stripe test payments and selected Supabase Auth workflows. Runs produce verification reports and resource state, with optional handover documentation; golive status checks recorded baselines for drift on demand. Changes go to your own provider accounts, and teardown is limited to resources GoLive can prove it created and requires separate approval.
After you install and invoke the golive skill in an app repository, the coding agent inspects the app and existing services, asks about missing providers, and presents the destination accounts and plan. CLI commands including detect, plan, apply, verify, status, handoff and teardown support detection, planning, execution, verification, drift checks, handover records and approved cleanup. apply requires a plan ID and --yes; DNS, live-mode and deletion actions require their respective additional confirmation flags. The CLI uses your signed-in provider accounts or local credentials, writes resources, and produces .golive/report.json, GOLIVE_REPORT.md and state files; golive handoff --write also creates GOLIVE_HANDOVER.md. A run stops at the first failed check or missing prerequisite and can resume from that step.
- A developer using Codex or Claude Code who wants to deploy an existing app to Vercel or Netlify.
- A team that wants to create a test database in its own Supabase or Neon account and connect it to an app.
- A maintainer attaching a custom domain to Vercel and approving DNS writes through Porkbun or GoDaddy.
- An app developer validating a Resend transactional-email setup or Stripe test-mode payment flow.
- A maintainer who needs resource ownership records, on-demand drift checks, or cleanup of temporary resources GoLive created.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
npx skills add https://github.com/mikehasa/golive-skill --skill golive --global --agent codex --yesnpx golive@alpha install --agent codex --globalnpx golive@alpha helpnpx golive@alpha detect
- Environment variables
GOLIVE_REPORT- Runtime
- Node.js 20+
How do you install or deploy this agent?
Requires Node.js 20+, npm/npx, Git, and a coding agent that can load skills and run commands. The README says installation has been checked for Codex and Claude Code; other clients are unverified. Install globally with the Skills CLI:
npx skills add https://github.com/mikehasa/golive-skill --skill golive --global --agent codex --yesFor Claude Code, replace --agent codex with --agent claude-code. The npm channel is another documented option:
npx golive@alpha install --agent codex --globalInstallation copies the skill but does not connect accounts or deploy. Omit --global to install for the current project.
How do you use this agent?
Open the target app repository in your coding agent. If GoLive does not appear, reload skills or start a new session. First invocation:
$golive Help me take this app live.In Claude Code, use /golive Help me take this app live.. Follow the prompts to sign in to required providers and review the destination accounts and plan. The npm package can also run individual CLI operations, for example:
npx golive@alpha help
npx golive@alpha detect
npx golive@alpha planTo make changes, supply the plan ID and explicit approval; DNS, production and deletion actions also need additional confirmations. Run verification afterward and inspect GOLIVE_REPORT.md; use golive handoff --write if you need the handover document.
What are this agent's strengths and limitations?
- Requires a reviewed plan before changes; writes need a plan ID and explicit approval, with extra gates for DNS, production and deletion.
- Supports Vercel/Netlify hosting and Supabase/Neon databases, with live-tested evidence for the documented paths.
- Produces inspectable verification reports, on-demand drift checks and ownership handover records; a failed step stops later steps.
- Can remove resources it can prove it created after approval, and lists resources it cannot remove as manual handover items.
- It is early alpha; first-account setup and all app frameworks are unvalidated, and some provider pairings have only mock coverage.
- Real deployments require your own cloud accounts, provider logins and infrastructure; a full launch may also involve DNS, database, email or payment services.
- Credentials are stored outside the repository in a local plaintext file with mode 0600, not an OS keychain; processes running as the same user can read it.
- Guided setup for other providers does not guarantee completion or equivalent verification, and some resources require manual removal.
How does this agent compare with similar options?
Compared with signing in to provider dashboards and configuring each service manually, GoLive connects inspection, planning, approval, execution and verification in a coding-agent workflow while leaving provider accounts in your control. For providers without built-in adapters, it offers best-effort guidance without the same completion or verification guarantees.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| GoLive Launch Assistant This agent | 81 · Good | Agent plugin / skillFree | ★ 1.6k | 1mo ago | TypeScript | Codex · Claude Code |
| IBM MCP Integrations | 34 · Major gaps | MCP server | ★ 421 | 1mo ago | — | — |
| OptimNow Cloud FinOps Library | 67 · Some gaps | Agent plugin / skillFree | ★ 75 | today | Python | Claude Code |
| Olares Personal Cloud OS | 66 · Some gaps | CLIFree | ★ 5.4k | 1d ago | Go | — |
How does FollowAgents rate this agent?
Why each dimension lost points
The source describes read-only detection, plan-before-write behavior, plan identity rechecks, and additional confirmation for DNS, destruction, and live actions. Adapter tests show credentials passed through Secret request headers and raw error bodies kept out of results. Data flow, credential file permissions, and leftover resources are explained. Least privilege and dependency security score 2: the evidence describes permission scopes and zero runtime dependencies, but does not establish that every adapter uses minimal permissions or that supply-chain risks are comprehensively addressed. Rollback is narrowly scoped and not live-validated, so it scores 2. Publisher identity is unknown and the packet does not establish clear ownership or attribution, so source attribution scores 1.
The CLI specifies one-document JSON output and exit codes, and the workflow stops on failed steps. README, SKILL, and source code support the workflow description. Scores are 2 because the supplied German README excerpt is truncated, preventing a full consistency check, and the evidence does not show actionable recovery guidance for every failure class. Dependency availability scores 2: runtime dependencies are zero and Node 20+ is specified, but provider services, installation channels, and external provider CLIs/APIs remain prerequisites.
The product targets deployment of Agent-built applications on user-owned accounts and covers hosting, databases, auth, email, and payments. It distinguishes automated adapters from best-effort guided work and states unverified environments and capability limits. Audience and boundaries score 3. Trigger precision scores 2 because the description covers a broad deployment goal and many related configuration requests. Environment fit scores 2: Codex and Claude Code are identified as verified clients, while other clients are unverified.
Documentation is organized across architecture, trust, recovery, providers, validation, and distribution. Installation steps, commands, examples, naming, and limitations are clearly described; the MIT license file matches package metadata. Examples and FAQ score 2 because examples are extensive but the supplied evidence does not show a complete FAQ. Versioning and changelog score 2: the alpha version and update behavior are documented, but a complete change history is not supplied. Maintenance responsibility scores 1: an author name and contribution guide exist, but no maintenance commitment, support channel, or response arrangement is established. Unverified publisher identity is not itself a deduction.
Each command emits one JSON document, and plans, verification, status, handoff reports, and teardown provide usable outputs for Agents and users. The local CLI and provider checks address the concrete work of moving an app from code to deployment, supporting a high marginal-value score. Cost-benefit scores 2: runtime dependencies are zero and no GoLive backend is needed, but Node, external installation tools, provider accounts or CLIs, and human steps are still required.
The documentation separates live-validated behavior, mock coverage, and roadmap items, with named steps, limits, and some observed outcomes. Source code and test examples corroborate credential handling and CLI behavior. Claim traceability and fact-inference separation score 3. Cross-source corroboration scores 2 because README, SKILL, source, and tests provide corroboration, but the packet is only a sample and the README excerpt is truncated, so not all material claims can be cross-checked.
- The packet is a repository sample and the German README excerpt is truncated; unseen files remain unknown and should not be treated as absent.
- The documentation says first production deployment needs extra confirmation. An Agent already logged in to a provider may still be able to write to that account outside a GoLive plan.
- Credentials are stored in a local plaintext file with mode 0600 rather than a system keychain. Some rollback behavior is not live-validated, and some provider resources require manual cleanup.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.1.0-alpha.9 GoLive 0.1.0-alpha.9
Release notes excerpt
GoLive 0.1.0-alpha.9 Version bump in package.json, .claude-plugin/marketplace.json, the READMEs and docs/DISTRIBUTION.md; the bundle is built with GOLIVE_RELEASE_REF=v0.1.0-alpha.9. The attached tarball is npm pack from the tagged commit. What ships - The UptimeRobot adapter (#77): UptimeRobot joins PostHog and Sentry as an automated monitoring provider — monitor select/adopt/create with a creation marker, the read-only uptime-monitor check that passes only when the linked monitor watches the production URL and reports up, and teardown through the API. …View on GitHub ↗ -
v0.1.0-alpha.8 GoLive 0.1.0-alpha.8
Release notes excerpt
GoLive 0.1.0-alpha.8 Version bump in package.json, .claude-plugin/marketplace.json, the READMEs and docs/DISTRIBUTION.md; the bundle is built with GOLIVE_RELEASE_REF=v0.1.0-alpha.8. The attached tarball is npm pack from the tagged commit. What ships - site-metadata (#72): the read-only launch-metadata check — title, description, canonical link and the Open Graph/Twitter share tags, warn-only, never failing. …View on GitHub ↗ -
Release notes excerpt
GoLive 0.1.0-alpha.7 — an alpha release of the skill and the bundled zero-dependency runtime. What changed since 0.1.0-alpha.6 - The ClawHub channel now installs a copy that verifies. 0.1.0-alpha.6 was published there, and the copy the registry installs still failed golive's own integrity check — ClawHub's review said the same. Comparing the published artifact file by file against the release manifest showed why: nineteen files matched and references/.gitkeep was missing. The registry's CLI skips every path with a dot-prefixed segment, so that empty placeholder could never be pu …View on GitHub ↗
FAQ
Do I need to pay for a GoLive service?
Can the agent change a production account without my approval?
apply requires a plan ID, --yes and any confirmation flag required for that action. However, an agent already signed in to a provider can bypass GoLive and write there directly.Which hosting and database providers are supported?
What happens if a step fails?
apply resumes at that step. Some teardown actions cannot be automated and are reported as handover tasks.Where are credentials stored?
~/.config/golive/credentials, a local plaintext file outside the repository with mode 0600, not an OS keychain. Values are kept out of plans, state, reports, command arguments and output.Is GoLive Launch Assistant open source?
Is GoLive Launch Assistant safe to use?
Is GoLive Launch Assistant still maintained?
Who makes GoLive Launch Assistant?
What is GoLive Launch Assistant's official website?
Thanks — we use this to improve the page.