@e9n/pi-context

Visual context window usage command for pi — see how your context budget is being spent, inspired by Claude Code's /context command.

Features

• Usage bar — colored hexagon grid showing filled vs free context
• Category breakdown — system prompt, tools, agents, skills, messages shown with distinct colors
• Per-item details — individual token counts for every tool, agent profile, and skill
• Autocompact buffer — shows reserved space for compaction headroom

Usage

/context

Example output:

Context Usage

⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛶ ⛶   claude-sonnet-4-5 · 38k/200k tokens (19.0%)
⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶
⛁ ⛁ ⛁ ⛁ ⛀ ⛶ ⛶ ⛶ ⛶ ⛶   Estimated usage by category
⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛝ ⛝ ⛝ ⛝
                           ⛁ System prompt: 3.2k tokens (1.6%)
                           ⛁ Tools: 18.4k tokens (9.2%)
                           ⛁ Custom agents: 1.5k tokens (0.8%)
                           ⛁ Skills: 1.1k tokens (0.6%)
                           ⛀ Messages: 14k tokens (7.0%)
                           ⛶ Free space: 129k (64.5%)
                           ⛝ Autocompact buffer: 33k tokens (16.5%)

Tools (12)
└ subagent: 2.8k tokens
└ bash: 1.4k tokens
└ read: 1.2k tokens
...

Color Legend

| Symbol | Color | Category | |--------|-------|----------| | ⛁ | Blue | System prompt | | ⛁ | Cyan | Tools | | ⛁ | Magenta | Custom agents | | ⛁ | Yellow | Skills | | ⛀ | Green | Messages | | ⛶ | Gray | Free space | | ⛝ | Red | Autocompact buffer |

Notes

• Autocompact buffer is based on pi's compaction.reserveTokens setting (default: 16,384 tokens). The value is read from your settings if configured, otherwise the default is used. See pi compaction docs for details.
• Category estimates are approximate — token counts use a chars/4 heuristic and may not exactly match the API-reported usage.

Install

pi install npm:@e9n/pi-context

License

MIT