context-doctor 🩺

See what's eating your LLM context window — and fix it.

Every long-running LLM conversation slowly fills up with junk: duplicated documents, 10k-token tool outputs nobody reads again, base64 blobs, stale history. You pay for those tokens on every single call, and model quality drops as the window fills.

context-doctor is a zero-config profiler + optimizer for LLM contexts. It works with Claude, GPT, Gemini message formats, and plugs into Claude Desktop, ChatGPT (developer mode), Cursor, Claude Code — any MCP-capable app — or runs standalone from the terminal.

Where the tokens go
────────────────────────────────────────────────────────
Tool results      ████████████████░░░░░░░░░░░░  57%  ~41k
System prompt     ██████░░░░░░░░░░░░░░░░░░░░░░  21%  ~15k
Assistant replies ████░░░░░░░░░░░░░░░░░░░░░░░░  13%  ~9.4k
User messages     ██░░░░░░░░░░░░░░░░░░░░░░░░░░   9%  ~6.5k

Findings (4)
────────────────────────────────────────────────────────
✖ Message #12 contains a base64/binary blob (~8.2k tokens). [save ~7.4k]
   → Never put base64 in text content — use the provider's file/image APIs.
▲ Tool result at message #7 (web_search) is ~6.1k tokens. [save ~4.9k]
   → Truncate or summarize large tool outputs before they enter history.

Quick start (30 seconds)

No install needed:

npx context-doctor analyze conversation.json --model claude-sonnet-5

Apply the safe fixes:

npx context-doctor optimize conversation.json --out slimmed.json

Input is any of: OpenAI chat format, Anthropic messages format (with system and content blocks), or a bare [{role, content}] array. Use - to pipe from stdin.

Related MCP server: CostGate

Always-on: optimize every request automatically

Run the proxy and every Anthropic/OpenAI API call your apps make gets optimized in flight — no code changes:

npx context-doctor proxy

Then point your app or SDK at it:

export ANTHROPIC_BASE_URL=http://localhost:8787      # Anthropic SDKs / tools
export OPENAI_BASE_URL=http://localhost:8787/v1      # OpenAI SDKs / tools

The proxy dedupes repeated content, trims stale tool results, and strips base64 blobs from the message history of each request, then forwards it to the real API. Your API key passes through in headers untouched, streaming (SSE) works unchanged, and per-request savings are logged:

[context-doctor] POST /v1/messages → 200 | optimized 7.3k → 518 tokens (2 changes)

Because prompt caching matches byte-identical prefixes, deterministic strategies are chosen so repeated requests stay stable — but if you rely on aggressive cache prefixes, start with --strategy strip-base64 --strategy dedupe and add more as you verify.

Note on desktop chat apps: Claude Desktop and the ChatGPT app talk to their own backends — no tool can sit in that path. For those, use the MCP integration below and add a line to your custom instructions like: "When a conversation gets long or includes large pasted content, proactively use context-doctor's profile_context tool and tell me what to trim." The model will then invoke it on its own.

Use it inside your AI app (MCP)

context-doctor ships an MCP server, so the AI itself can profile and slim context on demand.

Claude Desktop — add to claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "context-doctor": {
      "command": "npx",
      "args": ["-y", "context-doctor-mcp"]
    }
  }
}

ChatGPT desktop (developer mode), Cursor, Claude Code (claude mcp add context-doctor -- npx -y context-doctor-mcp), and any other MCP client: same command, their config syntax.

Then just ask: "profile this conversation with context-doctor" or paste an exported chat and say "what's eating my context?"

MCP tools

Use it as a library

import { parseConversation, profileConversation, optimizeConversation } from "context-doctor";

const profile = profileConversation(parseConversation(chatJson), "claude-sonnet-5");
console.log(profile.totalTokens, profile.findings);

const { conversation, tokensBefore, tokensAfter } = optimizeConversation(chatJson, {
  strategies: ["dedupe", "trim-tool-results", "strip-base64"],
});

What it detects

• Oversized tool results — the #1 context killer in agent loops

• Duplicate content — the same doc/result pasted twice

• Repeated identical tool calls — a signal your agent forgot earlier results

• Base64 / binary blobs in text content

• Long history past the point where models track the middle

• Cache-hostile ordering — volatile content before stable content breaks prompt caching (Anthropic cache_control, OpenAI automatic prefix caching)

• Window pressure — usage % against the target model's real context window

What it fixes (deterministically — no LLM calls, no API keys)

Everything the optimizer does is inspectable: it prints exactly which messages changed and how many tokens each change saved.

Why token counts are "~"

Exact counts require each provider's private tokenizer. context-doctor uses a calibrated chars-per-token heuristic (denser for code/JSON) that lands within ~10% — plenty accurate for finding what's heavy and measuring savings, and it keeps the tool fully offline with zero configuration.

Roadmap

• LLM-powered summarization strategy (bring your own key) for prune-history stubs

• Proxy: per-route strategy config + response token accounting

• context-doctor watch — live profiling of an agent's JSONL trace

• Exact tokenizer adapters (tiktoken, Anthropic count-tokens API) as optional plugins

• Session import from Claude Code / Cursor transcript formats

Contributions welcome — this project is small on purpose. Open an issue before a big PR.

License

MIT