Use case: issue triage
Issue triage and git history in a short reply
Context Mode Engine lets your agent read a full issue list, a pull request list or hundreds of commits and pass the model a short summary. On the Vite repository, 40 GitHub issues as JSON came to 178.7 KB, and the model read 0.8 KB.
Free · Local · Private · Forever
Context Mode Engine is source-available under the Elastic License 2.0. Free to use, for all time. It runs on your machine and keeps your data in ~/.context-mode.
Key facts
- What it is
- A use case for Context Mode Engine: GitHub issues, pull requests with their checks, and git history with file stats.
- Who it is for
- Maintainers and leads who ask an agent to sort issues, review open pull requests or learn the history of a large repository.
- Agents
- The Engine works in ten agents: Claude Code, Codex, Gemini CLI, Qwen Code, Copilot CLI, VS Code, Kiro, ZCode, OpenCode and Pi.
- Measured result
- 40 issues: 178.7 KB to 0.8 KB. 300 commits with file stats: 160.5 KB to 0.3 KB. Plugin benchmark of 2026-11-01, median of 5 runs, raw table.
- Scope
- Tool output in the agent where you install the Engine. Bytes that reach the model, per tool output.
The job
Monday morning, 40 new issues. You ask your agent to sort them by area, find the duplicates and pick the three to fix first. Each issue comes with its body, labels, comments and reactions. As JSON, the list is larger than most source files.
With Context Mode Engine, the list goes to a sandbox on your machine. The agent writes a script that groups the issues and counts them, and the model reads the result. The same holds for pull requests with their check runs, and for git log --stat over hundreds of commits.
What reached the model
| Tool output | Raw output | Reached the model | Sandbox time |
|---|---|---|---|
GitHub issues (gh issue list --json, vitejs/vite, 40 items) | 178.7 KB | 0.8 KB | 65 ms |
GitHub pull requests with checks (gh pr list --json, vitejs/vite, 25 items) | 190.9 KB | 0.8 KB | 79 ms |
Commit history with file stats (git log --stat, 300 commits, vitejs/vite) | 160.5 KB | 0.3 KB | 66 ms |
Plugin benchmark of 2026-11-01 on an Apple M1 Pro, default settings, median of 5 runs. Raw data: the result tables and the same file on GitHub. Rerun it with the steps on Benchmarks.
How it works
- The agent runs
ghorgitthroughctx_execute, and the output goes to the sandbox on your machine. - A short script groups, counts and sorts the items. The model reads its print.
- The full JSON stays in the local index. When you ask about one issue later, the agent searches for it and reads that issue.
Your GitHub token stays in your own gh setup. Read Security for the rules the sandbox follows.
Prompts to give your agent
List the open issues labelled bug. Group them by area, mark the likely duplicates, and tell me which three to fix first and why.
Read the last 300 commits with file stats. Which files change together most often, and which areas changed the most this month?
Check the open pull requests. Which ones have red checks, what is the first error in each, and which ones are ready to merge?