Provides tools for performing Git versioning operations such as commits, branching, and repository management through a prompt-driven agent.
Click on "Deploy 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., "@dispersl-mcp Generate a simple web app with Flask and SQLite"
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.
A Model Context Protocol (MCP) server implementation that integrates with Dispersl; The AI Dev Team, to give you multi-agents that work together to build software.
This MCP server can use other MCP servers as well for distributed, tool-driven workflows.
Built for modern AI-driven development, with support for multiple LLM models, multi-agent planning, and full SDLC automation.
Multi-agent orchestration (plan, code, test, git, docs, chat)
Code generation, test generation, and documentation
Git operations and repo management
Conversational agentic chat
API key and session management
Connect to and manage other MCP servers (local or remote)
Extensible with custom tools and external agents
Works with Cursor, VS Code, and other MCP-compatible clients
Related MCP server: Bernstein - Multi-agent orchestration
env DISPERSL_API_KEY=your-api-key npx -y dispersl-mcpnpm install -g dispersl-mcpFor the most up-to-date configuration instructions, see the Cursor MCP Server Configuration Guide.
Example Cursor MCP config:
{
"mcpServers": {
"dispersl-mcp": {
"command": "npx",
"args": ["-y", "dispersl-mcp"],
"env": {
"DISPERSL_API_KEY": "YOUR_API_KEY"
}
}
}
}export DISPERSL_API_KEY=your-api-key
npm run devYou can specify default models for each agent type:
export DISPERSL_PLAN_MODEL=deepseek/deepseek-chat-v3-0324:free
export DISPERSL_CODE_MODEL=anthropic/claude-sonnet-4
export DISPERSL_TEST_MODEL=anthropic/claude-sonnet-4
export DISPERSL_GIT_MODEL=meta-llama/llama-4-maverick:free
export DISPERSL_DOCS_MODEL=openai/gpt-4o-mini
export DISPERSL_CHAT_MODEL=openai/gpt-4o-miniTo connect to local and external MCP servers, use .dispersl/mcp.json:
{
"mcpServers": {
"anthropic-main": {
"command": "npx",
"args": [
"-y",
"--package=@anthropic-ai/mcp-server",
"anthropic-mcp-server"
],
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}"
}
},
"anthropic-backup": {
"command": "npx",
"args": [
"-y",
"--package=@anthropic-ai/mcp-server",
"anthropic-mcp-server"
],
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY_BACKUP}"
}
}
}
}Tool Name | Description |
list_models | List available models |
dispersl_code_agent | Generate code files and codebases based on a prompt using agentic execution |
dispersl_testing_agent | Generate end to end tests based on a prompt using agentic execution |
dispersl_git_agent | Execute codebase versioning operations with Git based on a prompt using agentic execution |
dispersl_new_docs_agent | Generate file by file technical documentation for a code repository using agentic execution |
dispersl_chat_agent | Chat with the Dispersl agent to get knowledge or insights about codebases using agentic execution |
dispersl_plan_agent | Multi-agent task dispersion using agentic execution (plan agent) |
start_session | Start a new agentic session |
end_session | End an active session |
add_mcp_server | Connect to an external MCP server and save to config |
remove_mcp_server | Disconnect from an MCP server and remove from config |
get_models | List available AI models |
get_keys | Get API keys for the authenticated user |
new_key | Generate new API key |
create_task | Create a new task |
edit_task | Edit a task by ID |
get_tasks | Get all tasks |
get_task | Get a task by ID |
cancel_task | Cancel a task by ID |
edit_step | Edit a step by ID |
get_steps | Get all steps |
get_step | Get a step by ID |
cancel_step | Cancel a step by ID |
get_usage_stats | Get usage stats |
get_language_stats | Get language usage stats |
get_agent_stats | Get agent query stats |
get_task_history | Get task history by ID |
get_step_history | Get step history by ID |
fetch_api_root | Fetch API root (utility endpoint) |
health_check | Health check endpoint |
This agent is able to interact with the user to fetch insights, shared memories and task progress to the user.
await client.callTool({
name: "start_session",
arguments: { session_id: "my-session" }
});
const response = await client.callTool({
name: "dispersl_chat_agent",
arguments: {
prompt: "Hello, how are you?",
model: "meta-llama/llama-4-maverick:free"
}
});
await client.callTool({
name: "end_session",
arguments: { session_id: "my-session" }
});This agent is able to coordinate all the agents i.e code, test, git, documentation to execute complex tasks. Agents work in sync and handover tasks to each other once they complete their assigned role.
const response = await client.callTool({
name: "dispersl_plan_agent",
arguments: {
prompt: "Plan a workflow for building and testing an ExpressJS web app using TypeScript",
model: "meta-llama/llama-4-maverick:free",
agents: ["code", "test", "git", "docs"]
}
});This agent is able to run autonomously. It can also collaborate with the other agents i.e code, test, git, documentation to execute complex tasks. Agents work in sync and handover tasks to each other once they complete their assigned role.
const response = await client.callTool({
name: "dispersl_code_agent",
arguments: {
prompt: "Create a simple hello world function",
model: "meta-llama/llama-4-maverick:free"
}
});await client.callTool({
name: "add_mcp_server",
arguments: {
name: "my-server",
command: "node",
args: ["dist/server.js"],
env: { PORT: "8080" }
}
});# Install dependencies
npm install
# Run in development mode
npm run dev
# Build
npm run build
# Run tests (requires DISPERSL_API_KEY)
npm test
# Lint
npm run lint
# Format code
npm run formatContributions are welcome! Please submit a Pull Request.
MIT License - see LICENSE file for details
Connect to an external MCP server and save to config
Inputs: args (required), name (required), command (required), env.
Cancel a step by ID
Inputs: id (required).
Cancel a task by ID
Inputs: id (required).
Create a new task
Inputs: body (required).
Chat with the Dispersl agent to get knowledge or insights about codebases using agentic execution
Inputs: prompt (required), mcp, model, voice, +4 more.
Generate code files and codebases based on a prompt using agentic execution
Inputs: prompt (required), mcp, model, context, +2 more.
Execute codebase versioning operations with Git based on a prompt using agentic execution
Inputs: prompt (required), mcp, model, context, +2 more.
Generate file by file technical documentation for a code repository using agentic execution
Inputs: url (required), mcp, model, branch, +4 more.
Multi-agent task dispersion using agentic execution (plan agent). Agent choices can either be use 'code', 'test', 'git', 'docs' as the agent choices
Inputs: prompt (required), agent_choice (required), mcp, model, +4 more.
Generate end to end tests based on a prompt using agentic execution
Inputs: prompt (required), mcp, model, context, +2 more.
Edit a step by ID
Inputs: id (required), body (required).
Edit a task by ID
Inputs: id (required), body (required).
Fetch API root (utility endpoint)
Inputs: no named parameters.
Get agent query stats
Inputs: body.
Get API keys for the authenticated user
Inputs: no named parameters.
Get language usage stats
Inputs: body.
Get a step by ID
Inputs: id (required).
Get step history by ID
Inputs: id (required), body.
Get all steps
Inputs: no named parameters.
Get a task by ID
Inputs: id (required).
Get task history by ID
Inputs: id (required), body.
Get all tasks
Inputs: no named parameters.
Get usage stats
Inputs: body.
Health check endpoint
Inputs: no named parameters.
List available models
Inputs: no named parameters.
Generate new API key
Inputs: no named parameters.
Disconnect from an MCP server and remove from config
Inputs: name (required).
Start a new agentic session
Inputs: session_id.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 28 tools
Most tools target distinct resources and actions, especially task/step CRUD, stats, and specialized agents. Minor ambiguity exists between the plan agent and specialized agents, and history/current getters require attention to descriptions.
The set uses consistent snake_case and mostly follows a verb_noun pattern. Minor deviations like new_key vs create_task, fetch_api_root, and the dispersl_* agent prefixes are readable but not perfectly uniform.
28 tools is heavy for a single MCP server, spanning tasks, steps, stats, agents, sessions, MCP configuration, models, and API keys. While many are distinct, the surface is large enough that consolidation or clearer grouping would improve usability.
Core task, step, agent, and key operations are present, but lifecycle coverage has notable gaps. There is no create_step, no delete/revoke for tasks, steps, or keys, no list MCP servers, and no session termination.
The AI orchestration agent for modern software teams.
AI work orchestration for plans, tasks, teams, and coding-agent dispatch.
AI-powered spec-to-task decomposition and execution orchestration for coding agents.
Coordination hub for AI coding agents: message teammates, ask humans, audit every event.
