Flashlight

MCP Server that uses DeepSeek's 1M context window for whole-codebase code search.

How it works

Flashlight loads your entire codebase into DeepSeek's context, then uses LLM understanding to find relevant code — no embeddings, no keyword matching, just brute-force full-context search.

It relies on DeepSeek's prefix caching for repeat queries: as long as the same prefix (system instructions + base code) is sent, tokens are served from cache (¥0.02/million tokens vs ¥1/million tokens for miss).

For large projects exceeding the 1M token limit, Flashlight automatically shards the codebase by directory, queries all shards in parallel, and merges results.

Related MCP server: Acemcp

Setup

1. Install

npm install -g @1percentsync/flashlight

2. Get a DeepSeek API key

Get one at platform.deepseek.com.

3. Configure MCP

Add to your MCP client config:

Claude Code (~/.claude.json under mcpServers):

{
  "flashlight": {
    "command": "flashlight",
    "env": {
      "DEEPSEEK_API_KEY": "sk-..."
    }
  }
}

Flashlight detects the workspace from MCP roots when the client provides them. If roots are unavailable, it falls back to the server process working directory.

Usage

The server exposes a single tool search with parameters:

Output

Results are returned as code snippets — the matched line ranges with line numbers.

Configuration

Environment variables

Project-level config

Create .flashlight/config.json in the workspace root to customize file extensions per project:

{
  "ext_whitelist": [".mdx", ".astro"],
  "ext_whitelist_override": false
}

Priority: project config > FLASHLIGHT_EXT_WHITELIST env var > built-in defaults.

The config is read once at process start. Changes require restarting the agent environment.

How caching works

Flashlight relies on DeepSeek's prefix caching. On first query, it sends all code and saves a base snapshot. On subsequent queries:

1. Detect file changes against the saved base

2. If changed tokens exceed FLASHLIGHT_CHANGE_THRESHOLD (default 10%) — rebuild the base entirely

3. Otherwise — reuse the stored base text and append only changed files as incremental context

This ensures the prompt prefix stays stable across queries, maximizing cache hit rate.

Sharding (large projects)

When a project exceeds FLASHLIGHT_MAX_CONTEXT_TOKENS, Flashlight automatically:

1. Splits files by directory — tries the whole project first, then recursively splits by top-level directories until each group fits

2. Queries all shards in parallel

3. Merges and deduplicates results

Each shard maintains independent cache state. Shard boundaries only change when a shard overflows (split eagerly, merge lazily).

Logs

Logs are written to .flashlight/flashlight.log in the workspace root. Each query logs:

• Snapshot size and shard plan

• File change detection

• Per-shard query cache hit ratio

• Search results

Cost

With deepseek-v4-flash on a ~50K token codebase:

License

ISC