Skip to content

Tessa commands and runtime

This page documents the slash commands Tessa accepts, the three-layer memory model, the energy system, and the per-plan query limits.

Slash commands

Command Effect
/mute Disable voice output
/unmute Enable voice output
/voice Toggle voice on/off
/voice on Enable voice mode
/voice off Disable voice mode
/hat <name> Wear a specific hat (coder, architect, researcher, reviewer, security)
/hat list List available hats
/hat off Remove current hat
/image <path> Analyze an image file
/paste Analyze image from clipboard
/clear Clear conversation history
/sessions List previous sessions
/session new Start a new session
/session restore <N> Restore a previous session
quit / exit Exit the assistant

Memory

Tessa's memory persists across sessions in three layers.

Layer Scope Persisted
Session Current conversation Until you exit
Workspace Current project/Grid Across sessions (saved to workspace)
User All workspaces Across sessions and projects

Memory commands

These are natural-language prompts, not slash commands:

"Remember that auth-service uses PostgreSQL 15"
"What do you remember about auth-service?"
"Forget everything about the old API"
"What am I working on?"

Tessa decides whether a fact belongs in Session, Workspace, or User memory based on context. You can also ask explicitly: "Remember this at the workspace level."

Memory actions

The memory tool supports twelve actions. Tessa selects them from what you ask; you don't type the action names.

Action What It Does
remember Store a fact (scope: session, workspace, or user)
recall Search stored facts by topic
list Show all stored facts
forget Remove facts matching a topic
track Set the current task you're working on
status Show what you're currently working on
update Record progress on the current task
done Mark the current task complete
audit Review user-memory access history
export Export stored user-memory facts
set-language Set Tessa's preferred response language
clear Clear facts (alias of forget)

Energy system

Tessa runs on an energy system. Paid plans move between two states as you use her through each window.

State Behavior
Optimal Your full subscription quality, full depth of reasoning on hard problems
Conserving A lighter, faster model. Every capability still works, simpler responses on complex tasks

Both states run every Tessa capability. Energy changes how deeply Tessa reasons on hard problems, not which skills she can use; it never restricts which operations you can run.

Paid plans always keep a working assistant: when a window's budget is spent they floor at Conserving and are never blocked. Start-tier (free) users are different: when a free user's window budget is spent, the assistant enters a third state, Replenishing, and serves no responses until the next window opens, with a live recharge countdown shown.

The remaining energy is visible in the Energy Bar in the chat panel. Energy refills on a rolling window, and Tessa tells you when she is back to full.

Query limits by plan

Plan Per 6-hour window (enforced) Daily total (informational)
Start ~1 query 7
Visualize ~8 queries 35
Analyze Large, scales per node Large, scales per node
Fuse Large, scales per node Large, scales per node

Limits are enforced in 6-hour rolling windows, not as a single daily pool: the daily total is divided across four windows. When a window's budget is spent, energy refills as the next window opens. The window budget is measured in tokens, so the query counts above are approximate: a few large requests consume a window faster than several small ones. Your account energy panel shows the exact remaining budget.

No tier is unlimited. Analyze and Fuse carry large budgets that scale with your monitored node count, not infinite ones. When a window's budget is spent, the assistant drops to lighter responses (the Conserving state) and is never blocked, exactly as every paid tier does.

These figures apply to Grid subscriptions (Start, Visualize, Analyze, Fuse). Personal subscriptions are a separate axis: they carry energy levels (E5, E25, E50, E150) measured in monthly token allotments rather than per-window query caps. Your account energy panel shows the active limit for either.