ollim-bot registers nine slash commands in Discord. All commands are
scoped to DMs with the bot and restricted to the bot owner (resolved
via application_info() at startup). Non-owners receive “not authorized”.
Quick reference
| Command | Description | Parameters |
|---|
/clear | Clear conversation and start fresh | None |
/compact | Compress conversation context | instructions (optional) |
/cost | Show token usage for this session | None |
/fork | Start a forked conversation | topic (optional) |
/model | Switch the AI model | name (required) |
/thinking | Toggle extended thinking | enabled (required) |
/interrupt | Stop the current response | None |
/permissions | Set permission mode | mode (required) |
/ping-budget | View or set ping budget | capacity, refill_rate (both optional) |
Session commands
/clear
Resets the conversation to a blank state. This:
- Clears session-level tool approvals and cancels pending permission prompts
- Exits any active interactive fork
- Logs a “cleared” session event
- Drops the active client connection
- Deletes the persisted session ID
The next message starts a fresh session with no prior context.
/compact
Compresses conversation context using the Claude Agent SDK’s built-in compaction.
Useful when a long conversation starts hitting context limits.
| Parameter | Type | Required | Description |
|---|
instructions | string | No | Optional focus for the summary (e.g., “keep the routine config details”) |
15 turns · 3h 12m · 45k tokens compacted
Use instructions to tell the compactor what information matters most.
Without it, the SDK picks what to preserve automatically.
/cost
Returns the cumulative token usage and cost for the current session in USD.
Model commands
/model
Switches the AI model for subsequent messages. The change applies to both the main session and any active fork.
| Parameter | Type | Required | Choices |
|---|
name | choice | Yes | opus, sonnet, haiku |
/thinking
Toggles extended thinking (chain-of-thought reasoning). When enabled, sets max_thinking_tokens to 10,000.
| Parameter | Type | Required | Choices |
|---|
enabled | choice | Yes | on, off |
Changing thinking mode drops the active client connection.
If you are in a fork, the fork client is interrupted and fork state is cleared.
Fork commands
/fork
Starts an interactive fork — a branched conversation that does not affect the main session history.
| Parameter | Type | Required | Description |
|---|
topic | string | No | Optional topic to seed the forked conversation |
/interrupt
Stops the bot’s current response mid-stream. The command is silent — it defers
the interaction, interrupts any in-progress generation, then deletes the
deferred response.
If the bot is not actively generating a response, /interrupt does nothing.
Configuration commands
/permissions
Sets the permission mode that controls how the agent requests tool approval.
| Parameter | Type | Required | Choices |
|---|
mode | choice | Yes | dontAsk, default, acceptEdits, bypassPermissions |
| Mode | Behavior |
|---|
dontAsk | Suppresses all prompts. Routes as default to the SDK while blocking locally. |
default | The agent asks for approval before using tools. |
acceptEdits | The agent can make file edits without asking, but other tools still require approval. |
bypassPermissions | All tools are auto-approved. No permission prompts. |
/ping-budget
Views or configures the ping budget — a token-bucket
rate limiter that controls how often background forks can notify you.
| Parameter | Type | Required | Description |
|---|
capacity | integer | No | Maximum number of pings in the bucket |
refill_rate | integer | No | Minutes between refills (default: 90) |
View budget
Update budget
Run /ping-budget with no parameters to see the current budget status.
Provide capacity, refill_rate, or both to update the budget.
The response confirms the new settings.
Next steps
