swipl-mcp-server

# SWI-Prolog MCP Server | Package | License | Node | |---------|---------|------| | [@vpursuit/swipl-mcp-server](https://www.npmjs.com/package/@vpursuit/swipl-mcp-server) | BSD-3-Clause | ≥20.0.0 | An MCP server that lets tools-enabled LLMs work directly with SWI‑Prolog. It supports loading Prolog files, adding/removing facts and rules, listing symbols, and running queries with deterministic pagination and true engine backtracking. ## Installation ### Claude Code CLI ```bash claude mcp add swipl-mcp-server npx @vpursuit/swipl-mcp-server ``` ### Claude Desktop ```json { "mcpServers": { "swipl": { "command": "npx", "args": ["@vpursuit/swipl-mcp-server"] } } } ``` - MacOS: `~/Library/Application Support/Claude/claude_desktop_config.json` - Windows: `%APPDATA%/Claude/claude_desktop_config.json` ### Cline (VS Code Extension) ```json { "mcpServers": { "swipl-mcp-server": { "autoApprove": [], "disabled": false, "timeout": 60, "type": "stdio", "command": "npx", "args": ["@vpursuit/swipl-mcp-server"] } } } ``` Configure via Cline's MCP settings in VS Code. ### Codex ```toml [mcp_servers.swipl-mcp-server] transport = "stdio" enabled = true command = "npx" args = ["@vpursuit/swipl-mcp-server"] ``` Add to `~/.codex/config.toml` ### ... and many others may also work ## Requirements - Node.js ≥ 20.0.0 - SWI‑Prolog installed and available in PATH ## Configuration ### Environment Variables Configure timeouts and logging via environment variables: | Variable | Description | Default | |----------|-------------|---------| | `SWI_MCP_READY_TIMEOUT_MS` | Server startup timeout (ms) | `5000` | | `SWI_MCP_QUERY_TIMEOUT_MS` | Query execution timeout (ms) | `30000` | | `MCP_LOG_LEVEL` | Logging verbosity | `warn` | | `DEBUG` | Enable debug logging (`swipl-mcp-server`) | - | | `SWI_MCP_TRACE` | Low-level spawn/protocol trace | - | | `SWI_MCP_PROLOG_PATH` | Override Prolog server script path | - | ## Features ### MCP Prompts Prompts guide AI assistants to help you with Prolog programming, knowledge base building and query optimization. **How it works:** 1. You select a prompt (via `/swipl` command in Claude Code CLI) 2. The prompt guides the AI assistant on how to approach your Prolog task 3. The AI assistant helps you with expert knowledge and step-by-step guidance *Note: Other AI assistants may access and use these prompts differently depending on their MCP implementation.* In Claude Code CLI, these prompts are available as slash commands. Simply type `/swipl` to see all available commands. - **`prolog_init_expert`** - Initialize expert Prolog assistance mode - **`prolog_quick_reference`** - Get comprehensive server overview - **`prolog_analyze_knowledge_base`** - Analyze current knowledge base state - **`prolog_knowledge_base_builder`** - Build domain-specific knowledge bases - **`prolog_query_optimizer`** - Optimize Prolog queries for performance ### MCP Resources Access to knowledge base and server information: - **`prolog://knowledge_base/predicates`** - List all predicates in the knowledge base - **`prolog://knowledge_base/dump`** - Export complete knowledge base - **`reference://help`** - Usage guidelines and tips - **`reference://license`** - License information - **`reference://capabilities`** - Server capabilities (JSON) ### Tools Overview - **Core:** `help`, `license`, `capabilities` - **Knowledge base:** `knowledge_base_load`, `knowledge_base_assert`, `knowledge_base_retract`, `knowledge_base_dump` - **Query:** `query_start`, `query_startEngine`, `query_next`, `query_close` - **Symbols:** `symbols_list` ## Protocol (short) - Each request/response is a single-line Prolog term. - Requests are enveloped as `cmd(ID, Term)` and responses as `id(ID, Reply)`; bare terms remain accepted. ## Usage Examples ### How the AI Assistant Helps You When you ask questions about Prolog, the AI assistant uses these tools behind the scenes: **You ask:** "Load my family relationships file" **AI uses:** ```json knowledge_base_load { "filename": "~/.swipl-mcp-server/family.pl" } ``` **You see:** "I've loaded your family relationships file." **You ask:** "Who are Mary's parents?" **AI uses:** ```json query_start { "query": "parent(X, mary)" } ``` then `query_next()` **You see:** "Mary's parent is John." **You ask:** "Add the fact that John is the parent of Mary" **AI uses:** ```json knowledge_base_assert { "fact": "parent(john, mary)" } ``` **You see:** "I've added that John is Mary's parent." **You ask:** "Show me all fruits in the list [apple, banana, cherry]" **AI uses:** ```json query_startEngine { "query": "member(X, [apple, banana, cherry])" } ``` with multiple `query_next()` **You see:** "The fruits in your list are: apple, banana, and cherry." **You ask:** "What predicates do I have defined?" **AI uses:** `symbols_list()` **You see:** "You have these predicates: parent/2, male/1, female/1..." ### Behind the Scenes The AI assistant manages: - Loading and saving your Prolog knowledge bases - Running queries and collecting all solutions - Adding and removing facts as you request - Optimizing your Prolog code - Ensuring security (files only from ~/.swipl-mcp-server/) ## Safety & Security - **File Path Restrictions**: Only `~/.swipl-mcp-server/` directory allowed for file operations - **Dangerous Predicate Blocking**: Pre-execution detection of `shell()`, `system()`, `call()`, etc. - **Hybrid Validation**: library(sandbox) validation plus explicit security checks - **Session Management**: Mutual exclusion between modes; timeouts for protection ## Troubleshooting - SWI‑Prolog not found: `swipl --version` must work - Increase `SWI_MCP_READY_TIMEOUT_MS`/`SWI_MCP_QUERY_TIMEOUT_MS` if timing out - Use `SWI_MCP_PROLOG_PATH` to point at a custom `server.pl` ## Package Info - Dependencies: `@modelcontextprotocol/sdk`, `zod` - License: BSD‑3‑Clause - Repository: https://github.com/vpursuit/swipl-mcp-server ## Support File issues and feature requests on GitHub. --- Built with [SWI‑Prolog](https://www.swi-prolog.org/) and the [Model Context Protocol](https://modelcontextprotocol.io/).