Contexity shows heuristic value metrics after relevant tool calls by default — a brief summary of context savings, percentages, and proof level. You can disable these visible summaries without losing access to the underlying structured metric data. The data is always available for inspection; toggling display controls only what the agent host surfaces to you.
Toggle Metrics Display
Use the metrics subcommand to control whether the visible summary footer appears:
# Disable visible summaries
contexity metrics disable --project /path/to/project
# Enable visible summaries
contexity metrics enable --project /path/to/project
# Check current status
contexity metrics status --project /path/to/project
What Changes
Toggling metrics display affects only the user-facing footer — not the underlying metric data.
When disabled:
- Hosts do not show the user-facing metrics footer after tool calls or task closeout
- Structured metric data can still be returned in the MCP response for programmatic inspection
- You can explicitly ask Contexity or your agent host to surface metrics at any time
When enabled:
- Hosts show saved amounts and percentages after relevant tool calls and at closeout
- Hosts preserve the proof-level label so you know how confident the metric is
- Heuristic metrics stay clearly labeled as heuristic — they are never presented as exact measurements
Control from the Agent Host
Agents can also manage this setting through MCP tool calls without leaving their workflow:
| Tool | Purpose |
|---|
context_set_metrics_display | Enable or disable the metrics footer programmatically |
context_get_metrics_display | Read the current display setting |
This lets an agent suppress noisy output during a long multi-step task and re-enable metrics display before surfacing a final summary.
If metrics output feels noisy during active development, disable it and re-enable before reviewing a completed task. This keeps the signal-to-noise ratio high during implementation while still giving you a clear summary at the end.
To understand what heuristic metrics measure, how they are calculated, and why they carry a proof-level label, see Metrics and Proof.