Configuration
All configuration lives in ~/.delta/config.json (created on first run).
The setup screen edits it for you; editing the file directly is for fine-tuning.
The file
{ "version": 1, "provider": "delta-free", "model": "glm-4.7-flash", "apiKey": "delta-free", "models": { "delta-free": "glm-4.7-flash" }, "mcpServers": { "exa": { "url": "https://mcp.exa.ai/mcp" } }, "permissions": { "bash": "ask" }, "timeouts": { "requestMs": 600000, "idleMs": 180000, "bashMs": 120000 }, "maxTokens": 8192, "retries": 100, "acceptedFreeModels": true}Settings
| Key | Default | Meaning |
|---|---|---|
provider | — | Provider id (delta-free, anthropic, openai, …) |
model | — | Active model id |
apiKey | — | delta-free for the free tier, or your key |
mcpServers | { exa } | Remote MCP servers (url per server) |
permissions.bash | ask | ask or allow |
timeouts.requestMs | 600000 | Max time for one model call |
timeouts.idleMs | 180000 | Kill the call if the model stalls this long |
timeouts.bashMs | 120000 | Max time for a single bash tool call |
maxTokens | 8192 | Output token cap per model call |
retries | 100 | Auto-retry cap for transient failures (429/5xx) |
Environment variables
| Variable | Effect |
|---|---|
DELTA_FREE_BASE_URL | Point the free tier at a different proxy |
DELTA_BASE_URL | Fallback base URL for the free tier |
| Any provider key var | Used instead of the config value when set |
Scope
Run deltacode inside a project and it edits only that project. Run it from
your home directory and it can work across the whole computer — writes outside
the current folder always ask permission first.