Clear Context

To clear context is to wipe an AI agent's conversation history so the next request starts from an empty context window. The model keeps its system prompt, tools and project instructions, but forgets every message, file and tool result from the session before it.

What Clear Context Means

An LLM has no memory between requests. Every turn, the agent resends the whole conversation, and that conversation is all the model knows about what has happened. Clearing context throws that conversation away. The agent does not lose its configuration, but it no longer remembers the last task.

In Claude Code you clear context by typing /clear at the prompt. Other agents have their own command, often /clear or /new.

Clear vs. Compact

	Clear	Compact
What it keeps	Nothing from the conversation	A short summary of it
When to use it	The next task is unrelated	The same task continues, but the history is long
Cost of the next turn	Lowest	Low
Risk	The agent loses details it still needed	The summary drops a detail

Why Clear Context Between Tasks

→ Cost. Every token left in context is sent again, and paid for, on every turn of the next task.
→ Focus. Files and errors from the last task compete for the model's attention with the current one.
→ Predictability. A task that starts from a clean context behaves the same whether it runs first thing in the morning or after ten other tasks.

The trade-off is real: anything the agent learned that the next task needs has to live somewhere other than the conversation, such as a file, a project instructions file, or workspace memory.

Clear Context in AgentRQ

An always-on agent never restarts between tasks, so its context only grows. AgentRQ's clear context option sends /clear to the agent's terminal before it picks up a task, then waits a moment and hands the task over. Every task starts clean, and no one has to remember to type it. The blog post on clearing context before every task works through the token savings.