Input Token Limit

Get alerted when an AI agent's prompts grow past your token budget. Input Token Limit catches bloated prompts and runaway context over a window and opens a violation when it passes your limit.

After Run(Detective)Checked after the trace lands, so it records and alerts

Prompts tend to grow. A retrieval step starts pulling in twenty chunks instead of five, or a conversation history never gets trimmed. Input Token Limit watches how many tokens go into the model and opens a violation when prompts get bloated.

Runs On
After traces land, on an Hourly, Daily, or Weekly window
You Set
An input token limit, how it is combined (Avg by default), and how often to check.
Your Agent Sends
Nothing extra. Normal tracing records token counts.

The template starts at Avg of input tokens above 8000, checked Daily, with Warning severity. You can change every value.

What Happens

What Traccia seesResult
The window is at or under your limitNo violation.
The window goes over your limitA violation opens. With Warn or Block, the next governed run is warned or stopped.
It drops back under the limit in the same windowThe violation resolves on its own (Auto Resolved).

This runs after the trace is accepted, so the run that triggered it is never undone. Observe records the violation. Warn also warns the next governed run. Block stops the next governed run. Ingest never rejects a trace because of a policy.

Set It Up

  1. In the app, open Policies and click Create Policy.
  2. Choose Input Token Limit under After Run.
  3. Give the policy a name and pick its scope: the organization, a workspace, or one Agent ID.
  4. Leave Metric on Input Token Count. Pick an Aggregation: Avg for typical prompt size, Max for the single largest, Sum for total tokens sent. Set the Threshold in tokens and the Frequency.
  5. Pick a mode. Observe only records. Warn and Block act on the next governed run.
  6. Turn on Email Alerts if you want mail when a violation opens.
  7. On Review, run Simulate Last 7 Days to see how often it would have matched, then click Activate Policy.

Good To Know

  • Input tokens are usually the bigger part of the bill. A sudden jump often means retrieval or history is stuffing the prompt.
  • Input Token Limit looks at finished traces. It does not trim a prompt.
  • Pair it with Output Token Limit to watch both sides of the conversation.

Where You See Results

Open the policy's Violations tab for what needs attention and Decision Log for every check. See Violations for the full lifecycle.

Next Steps

Output Token Limit

Watch response length as well.

Retrieval Completeness

Check that retrieval returned enough, live.

Policies

Every template, plus Observe, Warn, and Block.

© 2026 Traccia.