Bunny CLI and MCP Server

Package: bunny-tools · Binary: bunny · Node: ≥20

Why bunny-tools

Bunny.net is a great CDN/storage/DNS/streaming platform but managing it from the terminal was DIY territory. bunny-tools fixes that with one binary covering Storage, CDN, DNS, Stream, and Edge Scripting - and a built-in MCP server so AI agents can drive every command:

• bunny deploy for static sites. Familiar ergonomics if you've used firebase deploy - walk public dir → SHA-cached diff → parallel upload → CDN purge, in one command.
• bunny.json is your single source of truth. Versioned in git. Every command honors it. JSON Schema published at unpkg.com/bunny-tools/schema/bunny.schema.json.
• AI-native via MCP. AI agents see the same surface you do - no separate plugin per agent. bunny install mcp registers it with Claude Code in one shot.
• Verified end-to-end. 185 unit tests + nightly drift detection against a real Bunny account. We catch Bunny API changes before they break your deploys.

Install

As a CLI

npm install -g bunny-tools

As an MCP server (recommended for AI workflows)

For Claude Code:

claude mcp add bunny-tools npx -y bunny-tools mcp

For Claude Desktop, add to claude_desktop_config.json:

{
  "mcpServers": {
    "bunny-tools": {
      "command": "npx",
      "args": ["-y", "bunny-tools", "mcp"]
    }
  }
}

Once installed, any Claude session can use 15 high-level tools (bunny.deploy, bunny.purge, zone/dns CRUD) plus 3 resources (bunny://manifest, bunny://agents, bunny://config/current) without needing prior context. See MCP server below for the full tool surface.

Quickstart

bunny init                   # interactive: auth + feature picker + bunny.json
bunny deploy --dry-run       # preview
bunny deploy                 # storage sync + CDN purge

bunny init walks you through:

1. Bunny account API key (skipped if already in env or keychain)
2. Feature multi-select - Storage+CDN, DNS, Stream, Magic Containers, Edge Scripting
3. Per-feature config (e.g. for Storage+CDN: public dir, storage zone, password, pull zone, purge strategy)

Non-interactive form for CI:

bunny init --non-interactive --features=storage \
  --api-key=$BUNNY_API_KEY \
  --storage-zone=my-app \
  --storage-password=$BUNNY_STORAGE_PASSWORD \
  --pull-zone=12345

Add --ci to also generate .github/workflows/bunny-deploy.yml for GitHub Actions.

Quickstart for AI agents

With the MCP server installed (see Install), Claude Code and Claude Desktop can drive every command. Drop into any project and try:

| What you say | What happens | | --- | --- | | "Use bunny-tools to set up CI/CD for this project" | Runs bunny init, writes .github/workflows/deploy.yml using the official action, and lists the secrets you need to add | | "Deploy this site to Bunny" | Reads existing bunny.json, runs bunny deploy (with dry-run preview first) | | "Purge the CDN cache for tag release-2026-04" | Calls bunny.purge with tag:release-2026-04 | | "Show me what's on my Bunny account" | Calls bunny.whoami + lists reachable zone counts | | "Add an A record for www → 1.2.3.4 on my Bunny DNS zone for example.com" | Resolves the zone id, calls bunny.dns_record_add |

For best results in a new project, drop a 2-line hint into the project's CLAUDE.md:

## Deploy
This project uses bunny-tools. Run `bunny init` for first-time setup, then `bunny deploy`.

This anchors Claude to bunny-tools for that project's deploy work.

Setup & Auth

| Command | Description | | --- | --- | | bunny init | Interactive setup - auth + feature picker + bunny.json | | bunny configure | Walkthrough - store credentials in a named profile | | bunny configure list | List all credential profiles + scopes (masked) | | bunny configure switch <name> | Set the active profile | | bunny configure remove <name> [scope] | Remove a profile, or a single scope inside one | | bunny use [alias] | Switch active alias from .bunnyrc (or list aliases) | | bunny whoami | Show current account context + reachable zone counts | | bunny docs [topic] | Open Bunny.net docs in browser |

Example

bunny configure --profile=staging
bunny configure switch staging

Deploy & Purge

| Command | Description | | --- | --- | | bunny deploy | Sync public dir to storage zone and purge CDN cache | | bunny purge <target> | Purge by URL, tag:<name>, or pullzone:<id> |

Example

bunny deploy --dry-run
bunny purge tag:release-2026-05

Storage (file operations)

| Command | Description | | --- | --- | | bunny storage upload <local> <remote> | Upload a single file to the active zone | | bunny storage download <remote> <local> | Download a single file from a zone | | bunny storage list [path] | List a storage-zone path | | bunny storage delete <path> | Delete a file or path (use --recursive for dirs) | | bunny storage sync <local-dir> | Mirror local dir to zone - SHA-cached diff, parallel upload |

Example

bunny storage sync ./dist --zone=my-app

Storage Zones

| Command | Description | | --- | --- | | bunny storagezone list | List storage zones | | bunny storagezone get <id\|name> | Get a storage zone | | bunny storagezone create <name> | Create a storage zone | | bunny storagezone update <id> --body=<json> | Update a storage zone (raw JSON body) | | bunny storagezone delete <id> | Delete a storage zone |

Pull Zones (CDN)

Tip: bunny cdn ... works everywhere bunny pullzone ... does - Bunny's dashboard calls these "CDN", so the alias is there for muscle memory. Canonical name follows Bunny's API (pullzone).

| Command | Description | | --- | --- | | bunny pullzone list (or bunny cdn list) | List pull zones | | bunny pullzone get <id> | Get a pull zone | | bunny pullzone create <name> | Create a pull zone | | bunny pullzone update <id> --body=<json> | Update a pull zone (raw JSON body) | | bunny pullzone delete <id> | Delete a pull zone | | bunny pullzone edgerule list <id> | List edge rules on a pull zone | | bunny pullzone edgerule add <id> --rule=<json> | Add an edge rule (raw JSON rule) | | bunny pullzone edgerule delete <id> <rule-id> | Delete an edge rule | | bunny pullzone hostname list <id> | List custom hostnames linked to a pull zone | | bunny pullzone hostname add <id> <hostname> | Link hostname + provision Let's Encrypt cert + ForceSSL. Idempotent. --no-force-ssl opts out of HTTP→HTTPS redirect | | bunny pullzone hostname remove <id> <hostname> | Unlink a custom hostname |

Example

bunny cdn edgerule list 12345

# Wire DNS to a pull zone (2 steps - `add` does link + cert + ForceSSL
# in one idempotent call; the DNS record points at the wired-up PZ):
bunny pullzone hostname add 5780316 example.com           # ~2-90s
bunny dns record add 783181 PULLZONE @ --pull-zone=5780316

# Or do everything in one shot via the atomic Connect Domain command:
bunny domain connect 5780316 example.com --dns-zone=783181

DNS

| Command | Description | | --- | --- | | bunny dns list | List DNS zones | | bunny dns get <id\|domain> | Get a DNS zone | | bunny dns create <domain> | Create a DNS zone | | bunny dns delete <id> | Delete a DNS zone | | bunny dns record list <zone> | List records on a zone | | bunny dns record add <zone> <type> <name> <value> | Add a record (positional args) | | bunny dns record update <zone> <record-id> --body=<json> | Update a record | | bunny dns record delete <zone> <record-id> | Delete a record |

Supported record types: standard A, AAAA, CNAME, TXT, MX, SRV, CAA, NS plus Bunny routing types REDIRECT, PULLZONE, PTR, SCRIPT. PULLZONE and SCRIPT need --link-name=<id> (the linked pull zone / script id). For PULLZONE you can use the convenience flag --pull-zone=<id> instead and the CLI will fill in both the value and link-name from the pull zone's metadata. (FLATTEN is documented in Bunny's OpenAPI spec but the live API rejects it; dropped from supported types - re-add when Bunny enables it server-side.)

Examples

# Standard A record
bunny dns record add 783181 A www 1.2.3.4 --ttl=300

# Redirect www → https://example.com
bunny dns record add 783181 REDIRECT www https://example.com

# Wire DNS to a pull zone - auto-fills value + link-name
bunny dns record add 783181 PULLZONE "" --pull-zone=5780316

# Or raw form if you already know the pz name
bunny dns record add 783181 PULLZONE "" my-pz-name --link-name=5780316

Stream

| Command | Description | | --- | --- | | bunny stream library list | List Stream libraries | | bunny stream library create <name> | Create a Stream library | | bunny stream library delete <id> | Delete a Stream library | | bunny stream video list <library> | List videos in a library | | bunny stream video upload <library> <file> | Upload a video to a library | | bunny stream video delete <library> <video-id> | Delete a video |

Example

bunny stream video upload 42 ./demo.mp4 --title="My demo"

Magic Containers

| Command | Description | | --- | --- | | bunny containers app list | List container apps | | bunny containers app delete <id> | Delete a container app |

containers app create is deferred to v0.2 - Bunny's v3 schema requires runtimeType + containerTemplates[] + autoScaling which the current CLI surface doesn't yet model. Manage creation via the Bunny dashboard for now.

Edge Scripting

| Command | Description | | --- | --- | | bunny scripting list | List edge scripts | | bunny scripting deploy <name> --code=<file> | Create or update a script (dual-mode) | | bunny scripting delete <id> | Delete a script |

Example

bunny scripting deploy my-router --code=./worker.js

Discovery & AI

| Command | Description | | --- | --- | | bunny manifest --pretty | Full registry as JSON (machine-readable surface) | | bunny manifest --names | One command name per line - handy for shell completion | | bunny <any-command> --help-json | Help for any command as JSON | | bunny mcp | Boot MCP stdio server (15 tools, 3 resources) |

See AGENTS.md for AI-agent guidance.

Global flags

Apply to any command:

| Flag | Effect | | --- | --- | | -c, --config <path> | Override bunny.json location | | --cwd <dir> | Run as if launched from this directory | | -e, --env <alias> | One-shot .bunnyrc alias | | -p, --profile <name> | One-shot credential profile |

Example

bunny --profile=staging deploy

Configuration

bunny.json (per-project, git-tracked):

{
  "$schema": "https://unpkg.com/bunny-tools/schema/bunny.schema.json",
  "deploy": {
    "publicDir": "dist",
    "ignore": ["bunny.json", ".bunnyrc", "**/.*", "**/node_modules/**"],
    "storageZone": "my-app",
    "region": "ny",
    "concurrency": 8,
    "pullZones": [{ "id": 12345, "purge": "all" }]
  }
}

.bunnyrc (per-developer aliases, gitignored):

{
  "default": "prod",
  "aliases": {
    "prod":    { "storageZone": "my-app",     "pullZones": [12345] },
    "staging": { "storageZone": "my-app-stg", "pullZones": [12346] }
  }
}

Multi-account profiles (~/.config/bunny-tools/credentials.json):

{
  "active": "default",
  "profiles": {
    "default": { "account": "...", "storage:my-app": "..." },
    "staging": { "account": "...", "storage:my-app-stg": "..." }
  }
}

Switch the active profile with bunny configure switch <name>, or one-shot with bunny --profile=<name> <cmd> or BUNNY_PROFILE=<name>.

Auth model

Four credential scopes (all use the AccessKey HTTP header):

• account - Account API key
• storage:<zone> - Storage zone password (per zone)
• stream:<lib> - Stream library API key (per library)
• database:<name> - Database access key

Resolver chain (per call site): --flag → scoped env (e.g. BUNNY_STORAGE_PASSWORD_MY_APP) → generic env (BUNNY_STORAGE_PASSWORD) → OS keychain (<profile>:<scope> keys) → ~/.config/bunny-tools/credentials.json → interactive prompt.

GitHub Actions

bunny init --ci generates a workflow that uses npm-install + bunny deploy:

- name: Install bunny-tools
  run: npm install -g bunny-tools
- name: Deploy
  env:
    BUNNY_API_KEY: ${{ secrets.BUNNY_API_KEY }}
    BUNNY_STORAGE_PASSWORD_MY_APP: ${{ secrets.BUNNY_STORAGE_PASSWORD_MY_APP }}
  run: bunny deploy

A composite action wrapper (bytekcorp/bunny-tools-deploy-action@v1) is on the v0.2 roadmap for tighter ergonomics. The npm-install form above is the canonical path for v0.1.

MCP server (AI integration)

Install steps are at the top under Install → As an MCP server. Once configured, the server exposes:

Tools (15): bunny.deploy, bunny.purge, bunny.init, bunny.manifest, bunny.whoami, storage zone CRUD, pull zone CRUD, DNS zone + record CRUD, plus bunny.run as an escape hatch for any CLI invocation.

Resources (3):

• bunny://manifest - full command registry as JSON
• bunny://agents - AGENTS.md (workflows, gotchas, conventions for AI agents)
• bunny://config/current - resolved config from bunny.json + active alias

Cross-project usage. With the MCP server installed, drop a 2-line hint into any new project's CLAUDE.md to anchor Claude to bunny-tools for that project:

## Deploy
This project uses bunny-tools. Run `bunny init` for first-time setup, then `bunny deploy`. See `bunny manifest --pretty` for the full command surface.

Development

git clone https://github.com/bytekcorp/bunny-tools
cd bunny-tools
npm ci
npm run build
npm test
npm run dev -- manifest --pretty

For end-to-end testing against a real Bunny account (drift detection), see docs/e2e-testing.md.

For release instructions, see docs/deployment-guide.md.

FAQ

Is this the official Bunny.net CLI? No. bunny-tools is a community CLI. Bunny.net publishes @bunny.net/cli, which focuses on Databases, Magic Containers, and Edge Scripts. bunny-tools covers a different surface (Storage, CDN, DNS, Stream, Edge Scripting) and ships an MCP server for AI agents.

Why the binary name bunny? Short and matches the brand. Note that @bunny.net/cli also installs a bunny binary, so installing both globally will collide - pick one or use npx for the other.

Can I use this in CI? Yes. bunny init --ci generates a .github/workflows/bunny-deploy.yml. All commands honor BUNNY_API_KEY and per-zone scoped env vars.

Does it work without Node.js? Currently npm-only (Node 20+). A standalone binary is on the v0.2 roadmap.

License

MIT - see LICENSE.