SA
SkillAvatars
Back to MCP Tools

titan-agent

data-ai MCP Server

TITAN — The Intelligent Task Automation Network. Autonomous AI agent framework with GPU VRAM orchestration, NVIDIA cuOpt/NIM integration, ~155 tools, 34 LLM providers, 15 channels, mesh networking, LiveKit voice, self-improvement, and a React dashboard. npm i -g titan-agent

byDjtony707
VerifiedInstall Ready
data-aidata-ai
3 views17 stars6 forksv6.1.0-alpha.43NOASSERTION
View on GitHubHomepage

Why This Matters

Discovered via github-topic:mcp and last synced 3mo ago.

VerifiedInstall Ready
Source
github-topic:mcp
Stars
17
Last synced
3mo ago
Install
Instructions detected

Install

1. Install the package

npx titan-agent

2. Add to claude_desktop_config.json

{
  "mcpServers": {
    "titan-agent": {
      "command": "npx",
      "args": [
        "titan-agent"
      ]
    }
  }
}

Config file location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) / %APPDATA%\Claude\claude_desktop_config.json (Windows)

82
Tools
0
Resources
0
Prompts
Standard I/O
Transport

Available Tools (82)

Python

Python/.NET

36

Model-agnostic

coder

Shell, filesystem, edit

Pipeline

Model Name

train_prepare

Curate training data from conversation history

botbuilder

Enterprise integration

OpenAI-compat

M2.7 (2.3T MoE, 200K ctx), M2.5, M1

4

**GLM-4.7**

12

**MiniMax M2.7**

Category

Examples

Partial

—

gateway

agent

Mode

Behavior

Tool

What It Does

train_deploy

Convert model to GGUF and import to Ollama

irc-framework

Any IRC network

1

**GLM-5**

7

**Nemotron 3 Nano 4B**

Layer

What it covers

Human-in-the-loop

—

doctor

config

autonomous

Full auto — executes all tools, auto-triggers deliberation, self-initiates goal subtasks

self_improve_start

Launch an improvement session targeting a specific area

Panel

What It Does

Provider

Type

A-

5.3s

8

**MiniMax M2**

14

**DeepSeek V3.2**

System

Storage

Feature

TITAN

CrewAI

LangGraph

TypeScript

TypeScript

Growing

348k

Limited

RL training (CISPO)

ElevenLabs

—

0

0

Multi-machine

—

Client

Client

Dashboard

WS Broadcast

Template

Tools

explorer

Web search, fetch, browse

browser

Browser pool, Stagehand

analyst

Memory, data analysis

supervised

Asks before dangerous operations, proposes next steps (default)

locked

Asks permission for every tool call

Area

What Gets Optimized

titan-qwen

Single-turn instruction → tool selection pairs

titan-agent

Multi-turn ChatML conversations with OpenAI function calling

self_improve_status

Check current session progress

self_improve_apply

Apply successful experiment results to live config

self_improve_history

View history of all improvement sessions and outcomes

train_start

Launch LoRA fine-tuning job on GPU

train_status

Monitor training progress

Channel

Library

grammY

Bot API with webhook support

Baileys

No official API needed

Webhooks

Real webhook-based adapter

matrix-js-sdk

Decentralized chat support

Native

Any locally installed model

Model

Score

5

**Qwen 3.5 35B**

B

4.9s

Command

Description

Latency

Type

6

**Nemotron 3 Nano 24B**

13

**DeepSeek V3.1 671B**

yaml

Scaffold a skill project from template

2

**Devstral Small 2**

9

**Nemotron 3 Super**

Contributor

Contribution

3

**Qwen3 Coder Next**

10

**Kimi K2.5**

15

**Gemini 3 Flash Preview**

Widget

What It Does

11

`src/eval/harness.ts`

48420

`src/utils/constants.ts`

287

`tests/` (vitest)

Thing

Count

19

`src/channels/*.ts` (minus base)

43

`ui/src/components/admin/*Panel.tsx`

Say

What lands

MIT

`LICENSE`

Browse all MCP servers