probe-search 0.4.8

AI Agent Context Engine — give your coding agent a brain beyond code

probe

Give your AI agent a brain beyond code.

probe is a CLI tool and MCP server that indexes your project's docs and code, then serves curated, reranked context to AI coding agents in milliseconds.

---

Contents

• Why probe
• Quick Start
• MCP Server Setup
• Indexing
• How It Works
• Example Output
• CLI Reference
• Configuration
• Data Handling

---

Why probe

AI coding agents often answer project questions by grepping, reading files one by one, and assembling context over several tool calls. probe gives them one ranked search across docs and code.

What you get:

• Claude Code and Codex plugins with bundled MCP server and usage skill
• Manual Claude Code and Codex setup with probe install
• MCP first-use indexing and incremental refresh-before-search
• Hybrid keyword + semantic retrieval across text files, code, docs, and PDFs
• Cross-encoder reranking with ZeroEntropy zerank-2
• Local index storage in .probe/
• MCP tools and resources for Claude Code, Codex, and other MCP-compatible agents

---

Quick Start

Claude Code plugin (recommended)

Get a free ZeroEntropy API key at https://dashboard.zeroentropy.dev. Then run inside Claude Code:

/plugin marketplace add https://github.com/zeroentropy-ai/probe.git
/plugin install probe@zeroentropy
/reload-plugins

Use the HTTPS URL above. The zeroentropy-ai/probe shorthand makes Claude Code clone over SSH, which requires a configured GitHub SSH key. The Claude Code slash command treats --sparse as part of the URL, so do not pass sparse checkout options there.

Claude Code asks for your ZeroEntropy API key during plugin install. The plugin starts probe with uvx --from probe-search==0.4.8 probe mcp, so Claude Code does not need a separate probe install.

If you use the claude plugin install shell command instead of the /plugin slash command, pass the plugin config explicitly. From a shell, sparse checkout options are supported:

claude plugin marketplace add https://github.com/zeroentropy-ai/probe.git --sparse .claude-plugin plugins
claude plugin install --config zeroentropy_api_key="$ZEROENTROPY_API_KEY" probe@zeroentropy

You can also ask probe to install and configure the plugin from a shell:

uv tool install probe-search
probe install --client claude --plugin --api-key "$ZEROENTROPY_API_KEY"

Codex plugin

Get a free ZeroEntropy API key at https://dashboard.zeroentropy.dev. Then run from a shell:

codex plugin marketplace add https://github.com/zeroentropy-ai/probe.git --sparse .agents/plugins --sparse plugins/probe-codex
codex plugin add probe@zeroentropy
export ZEROENTROPY_API_KEY="ze_xxx"

The Codex plugin starts probe with uvx --from probe-search==0.4.8 probe mcp. Keep ZEROENTROPY_API_KEY in your shell environment before starting Codex, or run the direct installer below to persist the key in Codex's MCP config.

For Codex auto-review workflows, pre-approve probe's MCP tools and allow ZeroEntropy network egress once:

uv tool install probe-search
probe install --client codex --approve-tools --allow-zeroentropy-network

This writes a narrow Codex config allowlist for ZeroEntropy API access and package downloads, then pre-approves the probe MCP tools so auto-review does not block each search.

You can also ask probe to install the Codex plugin and direct MCP entry from a shell:

uv tool install probe-search
probe install --client codex --plugin --api-key "$ZEROENTROPY_API_KEY" --approve-tools --allow-zeroentropy-network

Direct CLI/MCP setup

probe install registers probe directly as an MCP server. It does not install the Claude Code or Codex plugin unless you pass --plugin.

uv tool install probe-search
# or: pipx install probe-search

export ZEROENTROPY_API_KEY="ze_xxx"
probe install --client claude
probe install --client codex
claude mcp list
codex mcp list

Use uv tool install or pipx so the registered probe path stays valid. If you install probe in a temporary virtual environment, rerun probe install after recreating that environment.

For custom Codex setups, pass the Codex config directory and binary explicitly:

probe install --client codex --codex-home ~/.codex-work --codex-bin /path/to/codex
probe doctor --codex-home ~/.codex-work --codex-bin /path/to/codex

For CLI-only use:

pip install probe-search
export ZEROENTROPY_API_KEY="ze_xxx"
probe index .
probe search "how does authentication work"

Verify your setup

probe doctor
probe smoke

probe doctor checks your API key, Claude Code/Codex wiring, MCP registration, and local index health without printing secrets or uploading project content. probe smoke indexes a tiny sample project and confirms search works. Use probe smoke --current to validate the current repo, probe smoke --claude to validate Claude wiring, and probe smoke --codex to validate Codex wiring.

---

MCP Server Setup

The Claude Code and Codex plugins are the best defaults. For another MCP-compatible agent, add a .mcp.json file to your project root:

{
  "mcpServers": {
    "probe": {
      "command": "uvx",
      "args": ["--from", "probe-search", "probe", "mcp"],
      "env": {
        "ZEROENTROPY_API_KEY": "ze_xxx"
      }
    }
  }
}

Your agent gets four tools:

Tool	What it does
probe_search	Search docs and code with automatic refresh and reranking
probe_index	Index or re-index project files
probe_status	Show what's indexed
probe_read	Read a file, optionally with focused line ranges

It also gets MCP resources:

Resource	What it exposes
probe://status	Index status and provider configuration
probe://files	Indexed file list
probe://file/{path}	Read a project file by URL-encoded path

When Claude Code starts probe, probe uses CLAUDE_PROJECT_DIR as the project root. Other agents should start probe from the project root or set their MCP server working directory to the project root.

---

Indexing

probe indexes all text-like files and PDFs, not just a fixed extension list. Unknown extensions and extensionless files such as Makefile, Dockerfile, and local config files are indexed if they look like text.

File discovery respects nested .gitignore files. It also reads .ignore files with higher precedence than .gitignore, which lets you keep files out of Git while still letting probe index them. Use .probeignore for probe-specific exclusions. probe always skips .git/, .probe/, __pycache__/, .venv/, compiled Python files, and obvious binary artifacts.

In MCP mode, the first probe_search creates the local .probe/ index automatically. For CLI-only use, run probe index . once. After .probe/ exists, CLI and MCP searches check for added, changed, and deleted files, then refresh only affected chunks. Set PROBE_REFRESH_TTL=0 to check before every search, or PROBE_REFRESH_TTL=-1 to disable automatic refresh.

Chunks keep Markdown header paths, code symbol names, PDF page numbers, and line ranges so agents can cite focused source locations.

---

How It Works

1. Hybrid retrieval: each query uses semantic vector search and SQLite FTS5 keyword search.
2. Reranking: candidates are fused and reranked with ZeroEntropy zerank-2.
3. Context assembly: results are deduplicated, trimmed to your token budget, and returned with file, section, and line metadata.

---

Example Output

$ probe search "how does authentication work"

 Found 5 results (342 chunks searched)

 [0.94] docs/design/auth.md > Authentication > OAuth Flow
   We use PKCE-based OAuth 2.0 with Auth0 as the identity provider.
   The flow works as follows: 1) Client generates a code verifier...

 [0.87] src/auth/oauth.py:42-71 > class OAuthHandler
   class OAuthHandler:
       """Handles OAuth2 PKCE flow for web and mobile clients."""
       def __init__(self, client_id: str, redirect_uri: str):

 ------------------------------------------
 zembed-1 + zerank-2 | 1,847 tokens | 0.3s

For scripts and agents:

probe search "how does authentication work" --json
probe status --json
probe doctor --json
probe smoke --json

---

CLI Reference

Command	Description
probe index [paths...]	Index project files for semantic search
probe index --full	Force full re-index
probe install --client claude	Register probe as a user-scope MCP server in Claude Code
probe install --client codex	Register probe as an MCP server in Codex
probe install --client codex --plugin	Install the Codex plugin and register direct MCP
probe install --client codex --approve-tools --allow-zeroentropy-network	Configure Codex auto-review to allow probe tool calls and ZeroEntropy egress
probe install --client codex --codex-home PATH --codex-bin PATH	Install against a custom Codex config directory or binary
probe install --client both	Register probe in Claude Code and Codex
probe search "query"	Search project knowledge with natural language
probe search --top-k N	Limit number of results
probe search --type code	Filter by file type
probe search --no-rerank	Skip reranking
probe search --max-tokens N	Set result token budget
probe search --json	Emit machine-readable results with line ranges
probe status	Show index stats and model config
probe status --json	Emit machine-readable index status
probe list	List indexed files
probe config	Show current model configuration
probe init	Create local config from environment
probe doctor	Diagnose API key, Claude Code, Codex, MCP, and index setup
probe doctor --json	Emit machine-readable diagnostics
probe doctor --codex-home PATH --codex-bin PATH	Diagnose a custom Codex config directory or binary
probe smoke	Run an end-to-end indexing and search validation
probe smoke --current	Smoke-test the current project
probe smoke --claude	Smoke-test search and Claude wiring
probe smoke --codex	Smoke-test search and Codex wiring
probe smoke --codex --codex-home PATH --codex-bin PATH	Smoke-test custom Codex wiring
probe mcp	Start MCP server
probe uninstall --client claude	Unregister probe from Claude Code
probe uninstall --client codex	Unregister probe from Codex
probe uninstall --purge	Unregister probe and delete .probe/

---

Configuration

probe stores its index and config in .probe/ at your project root. Add .probe/ to .gitignore.

# .probe/config.yaml
providers:
  embedding:
    name: zeroentropy
    model: zembed-1
    dimensions: 1280
  reranker:
    name: zeroentropy
    model: zerank-2

---

Data Handling

Project files are chunked and stored locally in .probe/ with SQLite and numpy. During indexing and search, probe sends query text and chunk text to ZeroEntropy for embedding and reranking. It does not create a remote document index; .probe/ is the durable project index.

---

Links

• ZeroEntropy -- the AI retrieval engine powering probe
• API Dashboard -- get your API key
• Documentation -- ZeroEntropy API docs
• Models -- zembed-1 and zerank-2 details

---

License

MIT -- see LICENSE for details.