ClawHub

Ai Config Admin

Other

Manage AI configuration for OpenClaw, OpenCode, Codex CLI, and Claude Code. Use when the user wants to add/remove models or providers, switch default or agent models, replace or update OpenClaw/OpenCode/Codex/Claude config files, or convert natural-language / irregular JSON / TOML input into the exact config changes supported by the bundled scripts. Triggers include: openclaw.json, opencode.json, config.toml, auth.json, settings.json, Claude Code, Codex CLI, provider, model, default model, agent model, OPENAI_API_KEY, ANTHROPIC_AUTH_TOKEN, ANTHROPIC_BASE_URL.

Install

openclaw skills install ai-config-admin

AI Config Admin

Identify which target the user wants to modify first: OpenClaw, OpenCode, Codex CLI, or Claude Code. Then use the bundled script for that target.

Required rules

  • Follow this skill and the existing script capabilities exactly.
  • Do not manually edit config files to bypass script limitations.
  • If the request is outside current script support, say so clearly and explain what is missing.
  • Before every write, create a backup in the same directory as the original file.
  • Perform backup and write through the bundled script, not manual editing.
  • Do not echo secrets such as apiKey, OPENAI_API_KEY, ANTHROPIC_AUTH_TOKEN, tokens, or refresh tokens in replies.

Target files

  • OpenClaw: ~/.openclaw/openclaw.json
  • OpenCode: ~/.config/opencode/opencode.json
  • Codex CLI: ~/.codex/config.toml, ~/.codex/auth.json
  • Claude Code: ~/.claude/settings.json

Routing

  • OpenClaw requests: openclaw.json, default model, agent model, add/remove provider, add/remove model, memory search.
  • OpenCode requests: opencode.json, OpenCode config, replace full JSON.
  • Codex CLI requests: config.toml, auth.json, Codex CLI, model_provider, review_model, wire_api, requires_openai_auth.
  • Claude Code requests: settings.json, Claude Code, ANTHROPIC_AUTH_TOKEN, ANTHROPIC_BASE_URL, ANTHROPIC_MODEL, CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC, CLAUDE_CODE_ATTRIBUTION_HEADER.

Input handling

  • Accept natural language, semi-structured JSON, or mixed descriptions.
  • Convert user intent into explicit script parameters or complete JSON/TOML text before invoking scripts.
  • Ask follow-up questions only when critical parameters cannot be inferred.

OpenClaw

  • Infer supported arguments from user intent.
  • When fields are missing, prefer filling from the current config when safe.
  • Use add-model to add a model to an existing provider, or to create/update a provider when the user also supplies the required provider fields.
  • For a brand-new provider, add-model needs enough provider config to make it valid: at minimum --base-url and --api; use --api-key and --auth-header when appropriate.
  • Use add-openai-model only for OpenAI-compatible provider setup flows that require explicit apiKey and api handling in one step.

OpenCode

  • If the user provides complete JSON, replace the file directly.
  • Do not merge partial JSON.

Codex CLI

  • config.toml: infer structured parameters and call the script.
  • auth.json: write only when the user provides a complete JSON payload; default to full replacement.
  • Do not delete unknown TOML sections such as projects, notice, or tui.
  • If existing auth.json contains login-state fields, replace it only when the user explicitly asks for full replacement.

Claude Code

  • Full settings.json input may be replaced directly.
  • If the user provides only an {"env": {...}} fragment:
    • If ANTHROPIC_BASE_URL is present, replace the Claude-related env key set.
    • Otherwise, update only the explicitly provided Claude-related env keys.
  • Do not invent missing model fields.
  • Do not modify non-Claude env keys.

Commands

Use {baseDir} as the skill root.

OpenClaw

python3 {baseDir}/scripts/openclaw_config.py --help
python3 {baseDir}/scripts/openclaw_config.py --file ~/.openclaw/openclaw.json summary
python3 {baseDir}/scripts/openclaw_config.py --file ~/.openclaw/openclaw.json add-model \
  --provider-id minimax-cn \
  --model-id MiniMax-M2.7 \
  --name 'MiniMax M2.7' \
  --context-window 200000 \
  --max-tokens 8192 \
  --set-default

Supported operations:

  • summary
  • add-model [--provider-id ...] [--model-id ...] [--name ...] [--context-window ...] [--max-tokens ...] [--reasoning true|false] [--allowlist] [--set-default] [--input text|image ...] [--base-url ...] [--api-key ...] [--api ...] [--auth-header true|false]
  • add-openai-model ...
  • set-openai-provider [--base-url ...] [--api-key ...]
  • remove-model <provider/model>
  • set-memory-search on|off
  • set-default-model <modelId>
  • set-agent-model <agentId> <modelId>
  • remove-provider <providerId>

OpenCode

python3 {baseDir}/scripts/opencode_config.py --file ~/.config/opencode/opencode.json summary
python3 {baseDir}/scripts/opencode_config.py --file ~/.config/opencode/opencode.json replace-from-stdin <<'EOF'
{...full JSON...}
EOF

Supported operations:

  • summary
  • replace-from-stdin

Codex CLI

python3 {baseDir}/scripts/codex_config.py --help
python3 {baseDir}/scripts/codex_config.py --config-file ~/.codex/config.toml --auth-file ~/.codex/auth.json summary
python3 {baseDir}/scripts/codex_config.py --config-file ~/.codex/config.toml replace-config-from-stdin <<'EOF'
model_provider = "OpenAI"
model = "gpt-5.4"
EOF
python3 {baseDir}/scripts/codex_config.py --auth-file ~/.codex/auth.json replace-auth-from-stdin <<'EOF'
{"OPENAI_API_KEY":"sk-..."}
EOF

Supported operations:

  • summary
  • replace-config-from-stdin
  • replace-auth-from-stdin
  • set-openai-provider [--model-provider ...] [--model ...] [--review-model ...] [--reasoning-effort ...] [--disable-response-storage true|false] [--network-access ...] [--windows-wsl-setup-acknowledged true|false] [--model-context-window ...] [--model-auto-compact-token-limit ...] [--provider-name ...] [--base-url ...] [--wire-api ...] [--requires-openai-auth true|false]

Claude Code

python3 {baseDir}/scripts/claude_config.py --help
python3 {baseDir}/scripts/claude_config.py --file ~/.claude/settings.json summary
python3 {baseDir}/scripts/claude_config.py --file ~/.claude/settings.json set-env \
  --anthropic-auth-token 'sk-...' \
  --attribution-header '0'
python3 {baseDir}/scripts/claude_config.py --file ~/.claude/settings.json replace-env \
  --anthropic-base-url 'https://example.invalid' \
  --anthropic-auth-token 'sk-...' \
  --attribution-header '0'
python3 {baseDir}/scripts/claude_config.py --file ~/.claude/settings.json replace-from-stdin <<'EOF'
{"env":{"ANTHROPIC_BASE_URL":"https://example.invalid"}}
EOF

Supported operations:

  • summary
  • replace-from-stdin
  • set-env [--anthropic-auth-token ...] [--anthropic-base-url ...] [--anthropic-default-haiku-model ...] [--anthropic-default-opus-model ...] [--anthropic-default-sonnet-model ...] [--anthropic-model ...] [--api-timeout-ms ...] [--disable-nonessential-traffic ...] [--attribution-header ...]
  • replace-env [--anthropic-auth-token ...] [--anthropic-base-url ...] [--anthropic-default-haiku-model ...] [--anthropic-default-opus-model ...] [--anthropic-default-sonnet-model ...] [--anthropic-model ...] [--api-timeout-ms ...] [--disable-nonessential-traffic ...] [--attribution-header ...]

Defaults and safeguards

  • Do not write headers.User-Agent unless explicitly needed.
  • Do not switch the default model unless the user asks.
  • Do not change memorySearch unless the user asks.
  • Stop if the script reports validation or dependency errors.
  • For OpenCode, default to full replacement when complete JSON is provided.
  • For Claude Code, if ANTHROPIC_BASE_URL is present, rebuild the Claude-related env set instead of carrying forward stale provider-specific env/model keys.