Stop losing context to large outputs — an MCP server for Codex CLI.

Forked from mksglu/claude-context-mode.

Note: Codex CLI hooks are not yet released. Until they ship, routing is driven by AGENTS.md instructions instead of automatic interception. Hooks support will be added once available (#2109).

• Sandboxed execution — runs shell commands and code in 10 languages, returns only stdout
• FTS5 knowledge base — indexes output with BM25 ranking, Porter stemming, heading-aware chunking
• Intent-driven search — query indexed content across your session
• URL fetching — fetches and indexes web pages without flooding context

git clone https://github.com/Wreos/codex-context-mode.git ~/.codex/plugins/context-mode
cd ~/.codex/plugins/context-mode
./install.sh

install.sh does two things:

1. Adds the MCP server to ~/.codex/config.toml
2. Appends routing instructions to ~/.codex/AGENTS.md

Then restart Codex.

Add to ~/.codex/config.toml:

[mcp_servers.context-mode]
command = "sh"
args = ["/path/to/codex-context-mode/start.sh"]

Add the contents of AGENTS.md to your ~/.codex/AGENTS.md.

Instead of running shell commands directly, use the MCP tools:

# Instead of running find/grep and reading each file individually:
mcp__context-mode__batch_execute(
  commands: [{label: "ts files", command: "find . -name '*.ts' | head -20"}],
  queries: ["component files", "utility functions"]
)

• pre_tool_use hook — automatic interception once Codex ships hooks (#2109)
• Stats command
• Doctor command

Original plugin by Mert Koseoğlu — this is a Codex CLI port by Wreos.

MCP has become the standard way for AI agents to use external tools. But there is a tension at its core: every tool interaction fills the context window from both sides — definitions on the way in, raw output on the way out.

With 81+ tools active, 143K tokens (72%) get consumed before your first message. And then the tools start returning data. A single Playwright snapshot burns 56 KB. A gh issue list dumps 59 KB. Run a test suite, read a log file, fetch documentation — each response eats into what remains.

Code Mode showed that tool definitions can be compressed by 99.9%. Context Mode applies the same principle to tool outputs — processing them in sandboxes so only summaries reach the model.

Each execute call spawns an isolated subprocess with its own process boundary. Scripts can't access each other's memory or state. The subprocess runs your code, captures stdout, and only that stdout enters the conversation context. The raw data — log files, API responses, snapshots — never leaves the sandbox.

Twelve language runtimes are available: JavaScript, TypeScript, Python, Shell, Ruby, Go, Rust, PHP, Perl, R, Dart, and Elixir. Bun is auto-detected for 3-5x faster JS/TS execution.

Authenticated CLIs work through credential passthrough — gh, aws, gcloud, kubectl, docker inherit environment variables and config paths without exposing them to the conversation.

When output exceeds 5 KB and an intent is provided, Context Mode switches to intent-driven filtering: it indexes the full output into the knowledge base, searches for sections matching your intent, and returns only the relevant matches with a vocabulary of searchable terms for follow-up queries.

The index tool chunks markdown content by headings while keeping code blocks intact, then stores them in a SQLite FTS5 (Full-Text Search 5) virtual table. Search uses BM25 ranking — a probabilistic relevance algorithm that scores documents based on term frequency, inverse document frequency, and document length normalization. Porter stemming is applied at index time so "running", "runs", and "ran" match the same stem.

When you call search, it returns relevant content snippets focused around matching query terms — not full documents, not approximations, the actual indexed content with smart extraction around what you're looking for. fetch_and_index extends this to URLs: fetch, convert HTML to markdown, chunk, index. The raw page never enters context.

Search uses a three-layer fallback to handle typos, partial terms, and substring matches:

• Layer 1 — Porter stemming: Standard FTS5 MATCH with porter tokenizer. "caching" matches "cached", "caches", "cach".
• Layer 2 — Trigram substring: FTS5 trigram tokenizer matches partial strings. "useEff" finds "useEffect", "authenticat" finds "authentication".
• Layer 3 — Fuzzy correction: Levenshtein distance corrects typos before re-searching. "kuberntes" → "kubernetes", "autentication" → "authentication".

The searchWithFallback method cascades through all three layers and annotates results with matchLayer so you know which layer resolved the query.

Search results use intelligent extraction instead of truncation. Instead of returning the first N characters (which might miss the important part), Context Mode finds where your query terms appear in the content and returns windows around those matches. If your query is "authentication JWT token", you get the paragraphs where those terms actually appear — not an arbitrary prefix.

The search tool includes progressive throttling to prevent context flooding from excessive individual calls:

• Calls 1-3: Normal results (2 per query)
• Calls 4-8: Reduced results (1 per query) + warning
• Calls 9+: Blocked — redirects to batch_execute

This encourages batching queries via search(queries: ["q1", "q2", "q3"]) or batch_execute instead of making dozens of individual calls.

The stats tool tracks context consumption in real-time. Network I/O inside the sandbox is automatically tracked for JS/TS executions.

Without context-mode, 9.6MB of raw tool output would flood your context window. Instead, 9.6MB (99%) stayed in sandbox — saving ~2,457,600 tokens of context space.

When installed as a plugin, Context Mode includes a PreToolUse hook that automatically injects routing instructions into subagent (Task tool) prompts. Subagents learn to use batch_execute as their primary tool and search(queries: [...]) for follow-ups — without any manual configuration.

Bash subagents are automatically upgraded to general-purpose so they can access MCP tools. Without this, a subagent_type: "Bash" agent only has the Bash tool — it can't call batch_execute or search, and all raw output floods context.

Measured across real-world scenarios:

Playwright snapshot — 56.2 KB raw → 299 B context (99% saved) GitHub Issues (20) — 58.9 KB raw → 1.1 KB context (98% saved) Access log (500 requests) — 45.1 KB raw → 155 B context (100% saved) Context7 React docs — 5.9 KB raw → 261 B context (96% saved) Analytics CSV (500 rows) — 85.5 KB raw → 222 B context (100% saved) Git log (153 commits) — 11.6 KB raw → 107 B context (99% saved) Test output (30 suites) — 6.0 KB raw → 337 B context (95% saved) Repo research (subagent) — 986 KB raw → 62 KB context (94% saved, 5 calls vs 37)

Over a full session: 315 KB of raw output becomes 5.4 KB. Session time before slowdown goes from ~30 minutes to ~3 hours. Context remaining after 45 minutes: 99% instead of 60%.

Full benchmark data with 21 scenarios →

These prompts work out of the box. Run /context-mode:stats after each to see the savings.

Deep repo research — 5 calls, 62 KB context (raw: 986 KB, 94% saved)

Research https://github.com/modelcontextprotocol/servers — architecture, tech stack,
top contributors, open issues, and recent activity. Then run /context-mode:stats.

Git history analysis — 1 call, 5.6 KB context

Clone https://github.com/facebook/react and analyze the last 500 commits:
top contributors, commit frequency by month, and most changed files.
Then run /context-mode:stats.

Web scraping — 1 call, 3.2 KB context

Fetch the Hacker News front page, extract all posts with titles, scores,
and domains. Group by domain. Then run /context-mode:stats.

Large JSON API — 7.5 MB raw → 0.9 KB context (99% saved)

Create a local server that returns a 7.5 MB JSON with 20,000 records and a secret
hidden at index 13000. Fetch the endpoint, find the hidden record, and show me
exactly what's in it. Then run /context-mode:stats.

Documentation search — 2 calls, 1.8 KB context

Fetch the React useEffect docs, index them, and find the cleanup pattern
with code examples. Then run /context-mode:stats.

• Node.js 18+
• Codex CLI (or Claude Code) with MCP support
• Optional: Bun (auto-detected, 3-5x faster JS/TS)
• Optional: Dart SDK (for Dart/Flutter code execution)

git clone https://github.com/Wreos/codex-context-mode.git
cd codex-context-mode && npm install
npm test              # run tests
npm run test:all      # full suite

MIT