Zibby

The cloud platform for AI agents — built on Claude Code, Cursor, Codex, and Gemini. Compose real coding-agent CLIs into structured agents with Zod-validated handoff between nodes, then deploy them to the cloud. Vendor-neutral. JavaScript-first.

An Agent is a deployed automation — a graph of coding-agent CLI calls with Zod-typed handoff between nodes. Build and ship one with the zibby agent CLI.

            ┌──────────┐    ┌──────────┐    ┌──────────┐
trigger →   │  plan    │ →  │ implement│ →  │  verify  │ → result
            │ (claude) │    │ (cursor) │    │ (codex)  │
            └──────────┘    └──────────┘    └──────────┘
                 │               │               │
              Zod out         Zod out         Zod out

graph
  .addNode('plan',      { prompt, outputSchema: Plan,   agent: 'claude' })
  .addNode('implement', { prompt, outputSchema: Diff,   agent: 'cursor' })
  .addNode('verify',    { prompt, outputSchema: Result, agent: 'codex'  });

Try it now

Get started in seconds:

npm install -g @zibby/cli
zibby init                       # bare init (config + creds; no template)
zibby agent new my-agent         # scaffold a custom agent
zibby agent run my-agent         # run locally (one-shot)

zibby login
zibby agent deploy my-agent      # ship to cloud
zibby agent trigger <uuid>       # fire it
zibby agent logs <uuid> -t       # tail logs

Want a ready-made agent instead of a blank slate? Browse the Agent Marketplace and use -t:

zibby init -t browser-test-automation     # scaffold a marketplace template at init time
zibby template list                        # see what's available
zibby template add <name>                  # add a template later (overwrites = doubles as update)

→ Next: Install

What you get

• Multi-vendor — mix Claude Code + Codex + Cursor + Gemini in one agent. Anthropic will never help you orchestrate Codex; multi-vendor neutrality is structural.
• Schema-enforced handoff — Zod validates every node's output before the next runs. No prompt-stuffing.
• Run anywhere — local with hot reload, or cloud with Heroku-style bundles (~3s cold start).
• Session replay — every run lands as on-disk JSONL + artifacts. Re-run any node via --session <id> --node <name>.
• Cloud-native — SSE log streaming, dedicated egress IPs for firewalled GitLab / GitHub Enterprise / Salesforce.
• Hosted apps too — Managed Apps host open-source tools (Grafana, Open WebUI, Docmost, Uptime Kuma, Authentik, …) from a curated catalog, OR deploy anything else via natural-language goal-mode. Every instance ships with an autonomous agent-ops sidecar that handles health checks, self-healing, and upgrades.
• Drive it from your AI agent — @zibby/mcp-cli exposes deploy / trigger / logs / debug as MCP tools. Add one snippet to Claude Code, Cursor, Codex, or Gemini and they call Zibby directly from chat. See Use from your AI agent.

Two product surfaces

	Agents	Apps
Lifetime	Per trigger (seconds-minutes)	Long-lived
Surface	Graph of agent CLI calls	A whole open-source application
Billing	Per execution	Per minute, while running
Best for	"When ticket lands, classify it"	"Host Grafana for the team"

An Agent is a deployed automation. An App is a hosted open-source application. Pick by how long the thing needs to run — see Apps overview for the decision tree.

How it compares

	Zibby	Claude Code Agent Teams	Devin	Mastra / LangGraph / CrewAI
Wraps real coding-agent CLIs	✅ Claude + Codex + Cursor + Gemini	❌ Claude only	❌ proprietary agent	❌ wraps LLM APIs
Multi-vendor	✅	❌	❌	N/A
Cloud-hosted	✅	❌ local IDE	✅	varies
Schema-enforced handoff	✅ Zod	❌ free-form	proprietary	✅
Session replay	✅ on-disk JSONL	❌	partial	partial
Static egress IP	✅ addon	❌	❌	❌