protocol-mcp
v3.1.1
Published
Protocol MCP server for Meta-Cognitive Swarm - Persistent memory and state tracking
Maintainers
Readme
Protocol MCP Server
Memoria - Long-term Memory for Meta-Cognitive Swarm Architecture
A Model Context Protocol (MCP) server that provides persistent state tracking and blueprint management for AI agent workflows.
Features
- 📝 Blueprint Management - Track multi-step projects with dependencies
- 🕐 Timestamp Tracking - Full audit trail of all state changes
- 📊 History Logging - JSONL audit log for forensic analysis
- 🔗 Dependency Management - Define and validate step dependencies
- ⚡ Cross-platform - Works on Windows, Linux, and macOS
- 🚀 Zero configuration - Works via npx, no installation needed
Installation
Via npx (Recommended)
No installation needed! Used automatically via MCP config:
{
"mcpServers": {
"protocol": {
"command": "npx",
"args": ["-y", "@baxtercooper/protocol-mcp"]
}
}
}Global Installation
npm install -g @baxtercooper/protocol-mcpLocal Development
git clone https://github.com/BaxterCooper/agentic.git
cd agentic/protocol-mcp
npm install
npm run build
npm startMCP Tools
init_blueprint(goal, steps)
Initialize a new project blueprint.
Parameters:
goal(string): The high-level project goalsteps(string[]): List of atomic task descriptions
Returns: Confirmation message with timestamp
Example:
{
tool: "init_blueprint",
arguments: {
goal: "Add user authentication",
steps: [
"Design auth architecture",
"Implement JWT service",
"Add login endpoint",
"Add middleware"
]
}
}update_step(step_id, status, notes?, priority?, dependencies?)
Update a step's status and metadata.
Parameters:
step_id(number): The ID of the step to updatestatus(string): PENDING | IN_PROGRESS | DONE | FAILED | BLOCKEDnotes(string, optional): Notes about the updatepriority(number, optional): -1=low, 0=normal, 1=highdependencies(number[], optional): Step IDs that must complete first
Returns: Update confirmation with old → new status
Example:
{
tool: "update_step",
arguments: {
step_id: 1,
status: "DONE",
notes: "Completed architecture design"
}
}read_protocol(include_history?)
Read the current protocol state.
Parameters:
include_history(boolean, optional): Include step change history
Returns: JSON representation of the protocol
Example:
{
tool: "read_protocol",
arguments: {
include_history: false
}
}get_audit_trail(step_id?, limit?)
Query the audit trail for historical events.
Parameters:
step_id(number, optional): Filter by specific steplimit(number, optional): Max events to return (default: 50)
Returns: JSON array of audit events with timestamps
Example:
{
tool: "get_audit_trail",
arguments: {
limit: 20
}
}check_dependencies(step_id)
Validate if a step's dependencies are met.
Parameters:
step_id(number): The step ID to check
Returns: Status message indicating if ready to proceed
Example:
{
tool: "check_dependencies",
arguments: {
step_id: 3
}
}Data Model
Protocol State
{
project_goal: string;
blueprint: BlueprintStep[];
created_at: string; // ISO 8601
last_modified: string; // ISO 8601
}Blueprint Step
{
id: number;
task: string;
status: "PENDING" | "IN_PROGRESS" | "DONE" | "FAILED" | "BLOCKED";
priority: number; // -1=low, 0=normal, 1=high
dependencies: number[]; // Step IDs
created_at: string; // ISO 8601
updated_at: string; // ISO 8601
history: HistoryRecord[]; // Change log
notes?: string;
}History Record
{
timestamp: string; // ISO 8601
old_status: string;
new_status: string;
notes: string;
}File Storage
- protocol.json - Current state (human-readable)
- protocol_history.jsonl - Append-only audit log
Location: .claude/memory/ in the project directory
Usage with Meta-Cognitive Swarm
This MCP server is part of the Meta-Cognitive Swarm Architecture. It serves as the Memoria (long-term memory) component.
See: https://github.com/BaxterCooper/agentic
Development
Build
npm run buildDevelopment Mode
npm run dev # Watch modeTesting
# Start server
npm start
# In another terminal, test with MCP clientLicense
MIT License - see LICENSE file
Version
3.0.0
Author
BaxterCooper
