melange-cli

Melange CLI

Give your AI agent access to Melange, an on-device AI deployment platform by ZETIC. melange is designed agents-first: non-interactive safe, structured --json output, stable exit codes, and machine-actionable errors.

Install

One line on macOS or Linux — installs the CLI and the agent skill that drives it:

curl -fsSL https://raw.githubusercontent.com/zetic-ai/melange-cli/main/script/install.sh | sh

Restart your coding agent afterward so it can discover the skill. Re-run the same line any time to update both.

The installer downloads the release binary for your platform, verifies its SHA-256 checksum, and installs the skill for universal agents and Claude Code. When cosign is on your PATH it also verifies the release-workflow signature; pass | sh -s -- --require-signature to make that mandatory.

curl -fsSL https://raw.githubusercontent.com/zetic-ai/melange-cli/main/script/install.sh \
  | sh -s -- --version v1.2.3 --agent "universal claude-code codex"

brew install zetic-ai/tap/melange

npm install -g @zetic-ai/melange-cli

go install github.com/zetic-ai/melange-cli/cmd/melange@latest

# Install for universal agents and Claude Code
npx skills add zetic-ai/melange-cli --skill melange-cli \
  --agent universal claude-code --global --yes

# Or choose one or more supported agents interactively
npx skills add zetic-ai/melange-cli --skill melange-cli --global

Authentication

By default melange auth login opens a browser for OAuth (recommended, zoa_/zor_ stored in OS keyring and auto-refreshed). For CI/headless use a Personal Access Token (ztp_) from Melange Settings → Personal Access Tokens:

export MELANGE_API_KEY="ztp_your_personal_access_token"
# or
melange auth login --with-token < token.txt

Or store it interactively once:

melange auth login   # browser OAuth; falls back to PAT paste if browser unavailable

Check status

melange auth status

Use the CLI directly

A read-only path to confirm the install works. Nothing here creates anything or counts against a quota:

melange auth login                        # or export MELANGE_API_KEY
melange library list --search whisper     # browse the public model library
melange library view zetic/whisper-tiny   # one model in detail
melange repo list                         # your own repositories

Add --json (optionally with --jq) to any command for machine-readable output.

Identifiers

Three different identifiers appear in command arguments, and they are not interchangeable — passing a repository or display name where a MODEL_KEY or TARGET_ID belongs is the most common mistake:

They chain in that order:

melange model list -R zetic/whisper-tiny
melange model targets MODEL_KEY -R zetic/whisper-tiny
melange model download MODEL_KEY -R zetic/whisper-tiny --target TARGET_ID --yes

melange model download is billable, so it requires --yes when it cannot ask interactively.

Build with AI

What's supported

With the agent skill installed, your agent can:

• Search the public model library and inspect available model versions.
• Compare real device benchmarks, targets, and report availability.
• Create and manage repositories, imports, uploads, and model versions.
• Generate exact deployment guides for Android, iOS, and Flutter.
• Check authentication, usage, quotas, and plan-specific availability.

Quick start (example prompts)

Discover a public model and deploy

Find a computer vision model by Meta in the Melange model library,
review its real device benchmark results, and give me the iOS deployment code/guide.

Upload your own model

I want to upload model.pt2 with sample.npy to Melange.
Upload the model, monitor conversion, and report the final status without retrying implicitly.

Compare benchmarks

Compare Gemma4 with another llm available in Melange. Use only benchmark values returned by Melange.
Show throughput and peak memory for iPhone 16 and Galaxy S25 where available.

MCP server

The CLI ships a built-in MCP server: melange mcp serves 18 tools over stdio (17 over HTTP — upload_model needs the caller's local files) so MCP clients call Melange directly instead of shelling out. The stdio server reuses the CLI's credentials (MELANGE_API_KEY or melange auth login), resolved lazily on the first tool call. Install the CLI first (see Install) so melange is on your PATH, then register it with your client. The full tool catalog and per-transport details are in llms.txt.

Claude Code

claude mcp add melange -- melange mcp

Verify with claude mcp list — the entry should show ✔ Connected.

Claude Desktop

Add to claude_desktop_config.json (Settings → Developer → Edit Config; macOS: ~/Library/Application Support/Claude/claude_desktop_config.json, Windows: %APPDATA%\Claude\claude_desktop_config.json), then restart Claude Desktop:

{
  "mcpServers": {
    "melange": {
      "command": "melange",
      "args": ["mcp"]
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project (or ~/.cursor/mcp.json for all projects):

{
  "mcpServers": {
    "melange": {
      "command": "melange",
      "args": ["mcp"]
    }
  }
}

Remote (Streamable HTTP)

For remote agent clients, serve the Streamable HTTP transport. The server itself holds no credentials: every request must carry its own token as Authorization: Bearer <token>, so one deployment serves many callers.

The server speaks plain HTTP; terminate TLS in front of it (load balancer, reverse proxy, or ingress). The https:// client URLs below assume that.

melange mcp --transport http --listen 0.0.0.0:8080

Claude Code (PAT ztp_... or OAuth zoa_...):

claude mcp add --transport http melange https://your-host:8080/ \
  --header "Authorization: Bearer ztp_your_personal_access_token" # or zoa_... OAuth

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "melange": {
      "url": "https://your-host:8080/",
      "headers": {
        "Authorization": "Bearer ztp_your_personal_access_token" // or zoa_... OAuth
      }
    }
  }
}

Claude Desktop registers local stdio servers through claude_desktop_config.json (above); remote servers are added as custom connectors in claude.ai settings instead. See melange mcp --help for the HTTP deployment flags (--validate-tokens, --allowed-origins, --resource).

Update

Re-run the installer to update the CLI and the skill together:

curl -fsSL https://raw.githubusercontent.com/zetic-ai/melange-cli/main/script/install.sh | sh

If you installed the CLI another way, update it the same way you installed it — brew upgrade zetic-ai/tap/melange or npm update -g @zetic-ai/melange-cli — and update the skill with npx skills update melange-cli --global.

Documentation

• Command reference (generated): docs/reference/melange.md
• LLM/agent surface reference: llms.txt

License

Apache-2.0