pi-zai-mcp

Give pi agents Z.ai-powered web search, URL reading, repository reading, and vision tools through MCP without leaving a pi session. This is an unofficial community package, not an official Z.ai package.

What you get

pi-zai-mcp registers pi tools that bridge Z.ai MCP servers:

• Search the web with Z.ai Web Search MCP.
• Read URLs and convert pages to model-friendly Markdown/text with Z.ai Web Reader MCP.
• Inspect GitHub repositories through Zread search, file reading, and directory-structure tools.
• Analyze images and videos through Z.ai vision tools for UI screenshots, OCR, error screenshots, diagrams, charts, image understanding, and video understanding.

Generic MCP tools are available immediately. Server-specific z_ai_* wrapper tools are registered after tool discovery, either by calling z_ai_mcp_list_tools or by setting Z_AI_MCP_AUTO_DISCOVER=1.

Proof from a real discovery run

A live z_ai_mcp_list_tools run on 2026-05-11 reported these Z.ai MCP tools:

search: web_search_prime
reader: webReader
zread: search_doc, read_file, get_repo_structure
vision: ui_to_artifact, extract_text_from_screenshot, diagnose_error_screenshot,
        understand_technical_diagram, analyze_data_visualization, ui_diff_check,
        analyze_image, analyze_video

The extension turns discovered tools into pi tool names with this pattern:

z_ai_<server>_<mcp_tool_name>

Examples:

z_ai_search_web_search_prime
z_ai_reader_webReader
z_ai_zread_search_doc
z_ai_zread_get_repo_structure
z_ai_zread_read_file
z_ai_vision_extract_text_from_screenshot

Install

Install from npm:

pi install npm:pi-zai-mcp

Install from GitHub:

pi install https://github.com/fitchmultz/pi-zai-mcp

Try it without installing permanently:

export Z_AI_API_KEY="your_z_ai_api_key"
pi -e npm:pi-zai-mcp

Run from a local clone:

git clone https://github.com/fitchmultz/pi-zai-mcp.git
cd pi-zai-mcp
npm install
export Z_AI_API_KEY="your_z_ai_api_key"
pi -e .

Configure

Variable	Required	Default	Purpose
Z_AI_API_KEY / ZAI_API_KEY	Yes	none	Z.ai API key used for HTTP MCP Bearer auth and the vision stdio server.
Z_AI_MCP_SERVERS	No	all	Comma-separated subset of search,reader,zread,vision.
Z_AI_MCP_AUTO_DISCOVER	No	off	Set to 1, true, yes, or on to discover and register server-specific wrappers at extension startup.
Z_AI_MCP_TIMEOUT_MS	No	30000	Per-connection/tool-call timeout in milliseconds.
Z_AI_MODE	No	ZAI	Passed through to the vision MCP server.

Example: disable vision server access for a lighter setup.

export Z_AI_MCP_SERVERS=search,reader,zread

Use

Built-in generic tools:

• z_ai_mcp_list_tools — list configured Z.ai MCP tools and schemas; successful discovery also registers server-specific wrappers.
• z_ai_mcp_call_tool — call an exact MCP tool by server id and raw MCP tool name.

Typical flow:

1. Ask pi to call z_ai_mcp_list_tools for search, reader, zread, or vision.
2. Use a discovered wrapper such as z_ai_search_web_search_prime, or call the exact MCP tool through z_ai_mcp_call_tool.
3. Run /zai-mcp-status in interactive pi to inspect server connection, discovery, and wrapper registration status.

Large MCP outputs are truncated to pi's standard 50 KB / 2000 line limit. When truncation happens, the full output is saved to a temp file and the path is included in the tool result.

How it works

• search, reader, and zread use Z.ai Streamable HTTP MCP endpoints.
• vision uses the @z_ai/mcp-server stdio server. The package depends on @z_ai/mcp-server@0.1.4 and falls back to npx -y @z_ai/mcp-server@0.1.4 only if the local binary is unavailable.
• The extension registers generic tools synchronously so pi startup is fast.
• Server connections and tool discovery are lazy by default to avoid blocking pi startup on network or package-manager work.
• session_shutdown closes any opened MCP transports.

Security and data flow

• Pi extensions run with your local user permissions. Review code before installing any third-party pi package.
• The extension reads Z_AI_API_KEY or ZAI_API_KEY from the environment; it does not store credentials.
• HTTP MCP calls send the key as a Bearer token to Z.ai MCP endpoints.
• Vision calls start a local stdio MCP server and pass the key in that child process environment.
• Truncated full outputs are written under your OS temp directory, not this repo.

Verify this repo

npm install
npm run typecheck
npm audit --omit=dev
npm publish --dry-run

For install-path checks, use a temporary project so local .pi/settings.json changes do not affect another repo:

tmpdir="$(mktemp -d)"
cd "$tmpdir"
pi install -l /path/to/pi-zai-mcp

Current limits

• Requires a Z.ai API key and network access for real tool calls.
• Server-specific wrapper availability depends on live MCP discovery. Generic list/call tools remain available even when discovery fails.
• Upstream MCP schemas and tool names can change.
• Verification currently consists of TypeScript typechecking, npm audit, npm dry-run packing, and pi install smoke checks; there is no dedicated unit test suite yet.

Project map

extensions/zai-mcp.ts  # public pi package entrypoint
src/index.ts           # extension implementation
package.json           # npm + pi package manifest
CHANGELOG.md           # release notes