Pyke MCP Server

An MCP (Model Context Protocol) server for the Pyke logic programming inference engine. This server enables LLMs to perform logical reasoning through knowledge bases with facts, rules, and queries.

Features

• Fact Management: Add ground truth facts to the knowledge base

• Rule Definition: Define inference rules using forward chaining

• Query Execution: Check if facts can be derived from the knowledge base

• Goal Proving: Find all variable bindings that satisfy a goal

• Session Management: Support multiple independent knowledge bases

• Logic-LLM Compatible: Load programs in Logic-LLM format

Installation

Using pip

pip install git+https://github.com/evertrol/pyke3 pyke-mcp

From source

git clone https://github.com/NewJerseyStyle/pyke-mcp.git
cd pyke-mcp
pip install -e .

Dependencies

• Python 3.10+

• mcp>=1.2.0

• pyke>=1.1.1 (Install from source by pip install git+https://github.com/evertrol/pyke3 but the package name is pyke not pyke3)

Quick Start

Running the Server

# Using the installed command
pyke-mcp

# Or using Python
python -m pyke_mcp.server

Configuration for Claude Desktop

Add to your claude_desktop_config.json:

macOS/Linux:

{
  "mcpServers": {
    "pyke": {
      "command": "pyke-mcp"
    }
  }
}

Windows:

{
  "mcpServers": {
    "pyke": {
      "command": "pyke-mcp"
    }
  }
}

Or with explicit Python path:

{
  "mcpServers": {
    "pyke": {
      "command": "python",
      "args": ["-m", "pyke_mcp.server"]
    }
  }
}

Available Tools

add_fact

Add a fact to the knowledge base.

Predicate(arg1, arg2, ...)

Examples:

• Human(Socrates, True) - Socrates is human

• Parent(John, Mary) - John is a parent of Mary

add_rule

Add an inference rule using forward chaining.

Premise >>> Conclusion
Premise1 && Premise2 >>> Conclusion

Examples:

• Human($x, True) >>> Mortal($x, True) - All humans are mortal

• Parent($x, $y) && Parent($y, $z) >>> Grandparent($x, $z) - Grandparent relationship

add_facts_and_rules

Bulk add multiple facts and rules at once.

query

Check if a specific fact can be derived.

Predicate(Subject, ExpectedValue)

Example:

• Mortal(Socrates, True) - "Is Socrates mortal?"

prove_goal

Find all variable bindings that satisfy a goal.

Examples:

• Mortal($who, True) - "Who is mortal?"

• Parent($parent, Mary) - "Who are Mary's parents?"

get_program

Display the current knowledge base contents.

clear_program

Clear all facts and rules from the session.

load_logic_program

Load a complete logic program from formatted text (Logic-LLM compatible).

list_sessions

List all active sessions.

delete_session

Delete a specific session.

Usage Examples

Classic Syllogism

# Add facts
add_fact("Human(Socrates, True)")
add_fact("Human(Plato, True)")

# Add rule
add_rule("Human($x, True) >>> Mortal($x, True)")

# Query
query("Mortal(Socrates, True)")
# Result: True, Match: True

# Find all mortals
prove_goal("Mortal($who, True)")
# Bindings: who=Socrates, who=Plato

Family Relationships

# Facts
add_fact("Parent(Alice, Bob)")
add_fact("Parent(Bob, Charlie)")
add_fact("Parent(Bob, Diana)")

# Rules
add_rule("Parent($x, $y) && Parent($y, $z) >>> Grandparent($x, $z)")

# Find Alice's grandchildren
prove_goal("Grandparent(Alice, $grandchild)")
# Bindings: grandchild=Charlie, grandchild=Diana

Loading a Complete Program

program = """
Predicates:
Human(x, bool)
Mortal(x, bool)

Facts:
Human(Socrates, True)
Human(Aristotle, True)

Rules:
Human($x, True) >>> Mortal($x, True)

Query:
Mortal(Socrates, True)
"""

load_logic_program(program)

Logic Program Format

The server accepts programs in the Logic-LLM format:

Predicates:
PredicateName(arg1_type, arg2_type, ...)

Facts:
PredicateName(value1, value2, ...)

Rules:
Premise($var, value) >>> Conclusion($var, value)
Premise1($x, val) && Premise2($x, val) >>> Conclusion($x, val)

Query:
PredicateName(subject, expected_value)

Syntax Rules

1. Facts: Ground assertions without variables

   • Human(Socrates, True)

   • Age(John, 30)

2. Rules: Implications with variables (prefixed with $)

   • Human($x, True) >>> Mortal($x, True)

   • Multiple premises joined with &&

3. Variables: Prefixed with $

   • $x, $person, $value

4. Values: Can be boolean (True/False) or strings

   • True, False, Socrates, Blue

Session Management

The server supports multiple independent sessions:

# Create/use a session
add_fact("Human(Socrates, True)", session_id="philosophy")

# Use a different session
add_fact("Cat(Whiskers, True)", session_id="animals")

# List sessions
list_sessions()

# Delete a session
delete_session("philosophy")

Error Handling

The server provides clear error messages for:

• Invalid fact/rule syntax

• Query execution failures

• Missing Pyke installation

• Session not found

Based On

• Pyke Knowledge Engine

• Logic-LLM - Pyke solver implementation

• clingo-mcp - MCP server interface reference

License

MIT License