Uses Cloudflare Durable Objects for persistent session memory storage, enabling AI agents to maintain context across conversations.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@NullShot Typescript MCP TemplateCreate a session for my smart contract audit with initial context about Solidity security patterns"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Cortensor MCP
An MCP server that plugs Claude, Cursor, and other AI assistants directly into Cortensor's decentralized inference network. Session persistence, task orchestration, and validation tools included.
Why Cortensor MCP?
The Problem
AI agents lack:
Persistent memory across conversations
Access to decentralized inference networks
Structured task orchestration tools
Built-in validation and consistency checking
Our Solution
Cortensor MCP provides a unified interface that:
Persists session memory using Cloudflare Durable Objects
Connects AI assistants to Cortensor's decentralized network
Orchestrates tasks with structured routing (analyze, summarize, extract, generate, validate, decide, plan)
Validates outputs against rubrics and checks for logical consistency
Result: AI agents gain persistent memory, decentralized inference capabilities, and powerful validation tools—all through a single MCP server.
Creating a Persistent Session
Routing Tasks to Cortensor
Validating Content Against a Rubric
Features
Tool | Description |
session_create | Create or retrieve an agent session with persistent memory |
session_remember | Store information in session memory for later recall |
session_recall | Retrieve stored information from a session |
task_route | Route structured tasks to Cortensor with output formatting |
validate_rubric | Validate content against a weighted scoring rubric |
validate_consistency | Check content for contradictions and logical issues |
research_analyze | Analyze code, docs, or issues and extract actionable items |
research_summarize | Generate digest, changelog, or status reports from multiple items |
cortensor_status | Check Cortensor network connectivity and latency |
cortensor_infer | Direct inference request to Cortensor network |
Setup Guide
1. Clone and Install
2. Run Locally or Deploy
MCP Configuration
Add to your MCP settings (Roo Code / Cline / Claude Desktop):
Task Types
Type | Behavior |
| Identify patterns and insights |
| Concise summary, preserve key info |
| Extract data points and entities |
| Generate content from requirements |
| Check correctness, flag issues |
| Reasoned decision with explanation |
| Actionable steps with priorities |