Session Settings
Fine-tune how the agent reads your codebase and how much it thinks before answering. All settings below are configured directly from the chat with a / command and take effect immediately.
/dir — Link an External Project
By default the agent reads the project where you started the chat. With /dir (also /directory or /folder) you can point it to any other directory on your machine — useful when working across multiple repositories or monorepos at the same time.
When you set a path, the agent automatically indexes the new directory in the background so it can answer questions about that codebase straight away.
/dir <path> — Set external directory
/dir ../my-other-service /dir /home/user/projects/backend-api /dir . # reset to current project
/dir — Open interactive menu
/dir # interactive menu: # ▶ External Project Path: /home/user/projects/backend-api (Click to change) # ▶ Apply and return to chat # ▶ Cancel and return to chat
/urls — Add Documentation Links
Give the agent access to external documentation pages — library references, API specs, internal wikis, or any public URL. The agent fetches and indexes those pages so it can reference them when answering your questions.
/urls <url1> <url2> ... — Set URLs
Pass one or more URLs separated by spaces. The list replaces any previously configured URLs:
/urls https://docs.stripe.com/api https://nuxt.com/docs /urls https://docs.anthropic.com/en/api/getting-started
/urls — Open interactive menu
/urls # interactive menu: # ▶ Documentation URLs: https://docs.stripe.com/api (Click to change) # ▶ Apply and return to chat # ▶ Cancel and return to chat
/max_tokens — Limit Response Length
Controls the maximum number of tokens the agent can generate in a single response. Lower values produce shorter, more direct answers. Higher values allow the agent to write out longer explanations, full files, or detailed reports.
Also available as /max-tokens or /maxtokens.
/max_tokens <number> — Set limit
/max_tokens 1024 # short, focused answers /max_tokens 4096 # default — balanced /max_tokens 16000 # long documents, full file output
/max_tokens — Open interactive menu
/max_tokens # interactive menu: # ▶ Max Tokens: 4096 (Click to change) # ▶ Apply and return to chat # ▶ Cancel and return to chat
| Value | Good for |
|---|---|
512 – 1024 | Quick clarifications, short code snippets |
2048 – 4096 | Everyday coding tasks (default) |
8000 – 16000 | Full file rewrites, long technical reports |
/effort — Reasoning Depth
Adjusts how much the model "thinks" before writing an answer. Supported on reasoning-capable models (e.g. o1, o3, Claude with extended thinking). On models that do not support it, the setting is silently ignored.
Also available as /reasoning.
/effort <level> — Set reasoning depth
/effort low # fast, less deliberate /effort medium # balanced (recommended for most tasks) /effort high # slower, deeper reasoning for complex problems /effort none # disable extended thinking (provider manages) /effort default # reset to provider default
/effort — Open interactive menu
/effort # interactive menu: # ▶ Reasoning Effort: Default (None / Provider Managed) (Click to change) # ▶ Apply and return to chat # ▶ Cancel and return to chat
| Level | Speed | Best for |
|---|---|---|
low | Fastest | Autocomplete, simple lookups |
medium | Balanced | Code generation, refactors, bug fixes |
high | Slower | Architecture decisions, complex debugging, security analysis |
high when the quality of the answer matters more than speed.