Setup Wizard¶

The setup wizard creates ~/.missy/config.yaml with your provider credentials and security policies. It supports both an interactive mode (the default) and a non-interactive mode for scripting and CI.

Interactive mode¶

missy setup

Step 1: Workspace¶

The wizard prompts for a workspace directory. This is where Missy is allowed to read and write files.

Step 1 of 5 — Workspace
  Workspace directory [~/workspace]:

Press Enter to accept the default (~/workspace) or type a custom path. The directory is created automatically if it does not exist.

Step 2: Provider selection¶

Step 2 of 5 — AI Provider(s)
  Choose which providers to configure:

    1. Anthropic (Claude)
    2. OpenAI (GPT-4o / Codex)
    3. Ollama (local models)
    4. Anthropic + OpenAI (both)
    5. All three
    0. Skip (configure manually later)

You can configure multiple providers. The first one listed becomes the default.

Step 3: API key and authentication¶

The wizard supports several authentication methods depending on the provider.

AnthropicOpenAIOllama

Three authentication options:

Auth method:
  1. API key  (sk-ant-api…)  [recommended]
  2. API key + vault  (encrypted, config references vault://)
  3. Claude Code setup-token  (sk-ant-oat…)  [ToS risk]

Option 1 (API key) is the standard path. The wizard checks for the ANTHROPIC_API_KEY environment variable first:

Detected ANTHROPIC_API_KEY in environment: sk-ant…wxyz
Use this key? [Y/n]:

If the variable is set and you confirm, Missy uses it at runtime without embedding the key in the config file.

Option 2 (vault) encrypts the key using Missy's ChaCha20-Poly1305 vault and stores a vault://ANTHROPIC_API_KEY reference in the config.

Option 3: setup-tokens

Setup tokens (sk-ant-oat...) are issued by Claude Code and are not supported by the Anthropic API. The wizard warns about Terms of Service implications. Use a regular API key from console.anthropic.com instead.

Two authentication options:

Auth method:
  1. API key  (sk-…)
  2. OAuth / Codex CLI  (browser flow, PKCE)

Option 1 prompts for an API key, with environment variable detection for OPENAI_API_KEY.

Option 2 opens a browser-based OAuth flow (PKCE) for ChatGPT / Codex CLI access. This switches the provider to openai-codex, which routes through chatgpt.com/backend-api instead of api.openai.com.

Remote / headless OAuth

For servers without a display, forward port 1455: ssh -L 1455:localhost:1455 user@host, then run the wizard.

No API key required. The wizard prompts for:

Base URL (default: http://localhost:11434)
Model name (default: llama3)

Ollama runs locally, so no network policy entries are added.

After entering credentials, the wizard offers to verify connectivity with a test API call:

Verify API key with a test call? [Y/n]:
  Connecting…
  Connection successful.

Model tiers¶

For Anthropic and OpenAI, the wizard presents the available models and asks you to assign three tiers:

Available models:
  1. Sonnet 4.6 — balanced (recommended)  (claude-sonnet-4-6)
  2. Haiku 4.5 — fastest / cheapest  (claude-haiku-4-5-20251001)
  3. Opus 4.6 — most capable  (claude-opus-4-6)

  Primary model [1-3]: 1
  Fast (simple tasks) model [1-3]: 2
  Premium (complex tasks) model [1-3]: 3

Tier	Used for
Primary	Default model for `missy ask` and `missy run`
Fast	Simple tasks, internal lookups, cost-sensitive operations
Premium	Complex multi-step reasoning, code generation

Step 4: Discord (optional)¶

Step 4 of 5 — Discord Integration  (optional)
  Configure Discord bot? [y/N]:

If you choose yes, the wizard collects:

Bot token (hidden input, from discord.com/developers)
Application ID
DM policy: disabled, allowlist, pairing, or open
Guild policies: per-server settings for require-mention, allowed channels, and capability mode
Acknowledgement reaction (default: eyes)
Ignore bots toggle

Note

Discord setup adds discord.com and gateway.discord.gg to the network allowlist automatically.

Step 5: Review and write¶

The wizard displays a summary table and asks for confirmation:

┏━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Setting            ┃ Value                                     ┃
┡━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ Config path        │ /home/you/.missy/config.yaml              │
│ Workspace          │ /home/you/workspace                       │
│ Provider: anthropic│ claude-sonnet-4-6  auth=sk-ant…wxyz       │
│ Verified: anthropic│ OK                                        │
└────────────────────┴───────────────────────────────────────────-┘

  Write this configuration? [Y/n]:

On confirmation, the wizard:

Backs up any existing config file
Writes ~/.missy/config.yaml atomically (permissions: 0600)
Creates ~/.missy/secrets/, ~/.missy/logs/, and ~/.missy/jobs.json if missing

Non-interactive mode¶

For automated deployments, CI pipelines, or scripted setups, pass --no-prompt with the required options:

missy setup \
  --provider anthropic \
  --api-key-env ANTHROPIC_API_KEY \
  --no-prompt

Required flags¶

Flag	Description
`--provider NAME`	Provider name: `anthropic`, `openai`, `openai-codex`, or `ollama`
`--no-prompt`	Disable all interactive prompts

Authentication (one of)¶

Flag	Description
`--api-key-env VAR`	Name of environment variable containing the API key
`--api-key VALUE`	API key value directly (use with caution)

Avoid --api-key in scripts

Passing an API key on the command line exposes it in shell history and process listings. Prefer --api-key-env and set the variable in your environment.

Optional flags¶

Flag	Description	Default
`--model NAME`	Model identifier	Provider's primary model
`--workspace PATH`	Workspace directory	`~/workspace`

Examples¶

Anthropic from env varOpenAI with custom modelOllama (no key needed)CI / Docker

export ANTHROPIC_API_KEY="sk-ant-api03-..."
missy setup --provider anthropic --api-key-env ANTHROPIC_API_KEY --no-prompt

export OPENAI_API_KEY="sk-..."
missy setup \
  --provider openai \
  --api-key-env OPENAI_API_KEY \
  --model gpt-4-turbo \
  --no-prompt

missy setup --provider ollama --model llama3 --no-prompt

RUN pip install -e . && \
    missy setup \
      --provider anthropic \
      --api-key-env ANTHROPIC_API_KEY \
      --workspace /app/workspace \
      --no-prompt

The non-interactive mode writes the same config.yaml structure as the interactive wizard, with identical security defaults (network deny-all, shell disabled, plugins disabled).

Re-running the wizard¶

You can run missy setup again at any time. If a config file already exists, the wizard asks before overwriting and creates a timestamped backup of the previous config.

Editing config directly

After initial setup, most users edit ~/.missy/config.yaml directly. Missy supports hot-reload: changes take effect without restarting. See Configuration Reference for the full schema.