How To Use

This app is an MCP server named workflow-os.

It helps an MCP client route a request into one of the bundled workflow domains, run a single main workflow, optionally apply one utility pass, and end with one concrete next action.

What It Does

The server exposes:

• tools for routing and execution
• prompts for common workflow entry points
• resources that expose the bundled workflow markdown files

The intended pattern is:

1. Route the task first
2. Select or inspect the right workflow/resource if needed
3. Run one main workflow
4. Optionally apply one utility prompt
5. Return one immediate next action

Start The Server

Build first:

pnpm install
pnpm run build

Run in stdio mode:

node dist/index.js

Run in HTTP mode:

MCP_TRANSPORT=http MCP_PORT=<port> node dist/index.js
# Example: MCP_TRANSPORT=http MCP_PORT=3000 node dist/index.js

HTTP endpoints:

• GET /health
• GET /docs
• GET /docs-api.json
• POST /mcp
• GET /mcp
• DELETE /mcp

In HTTP mode, interactive API docs are available at /docs, and the raw OpenAPI document is available at /docs-api.json.

Environment Variables

Main Tools

route_task

Use this first.

Input:

• task
• optional goal
• optional audience
• optional constraints
• optional preferred_mode
• optional preferred_domain

Output:

• mode
• domain
• confidence
• reason
• sequence
• use_utility
• utility_candidates

Example intent:

Route: "Help me validate a digital product idea for busy founders"

select_domain_workflow

Use this when you want the workflow resource URI and normalized stage before running.

Input:

• domain
• optional stage

Output includes:

• resource_uri
• subsection_uri
• stage
• command_format
• response_shape

run_workflow_sequence

This is the primary execution tool.

Input:

• mode
• domain
• task
• optional context
• optional stage
• optional optimize_once
• optional next_action_required

Output includes:

• main_deliverable
• applied_sequence
• optimization_applied
• next_action

apply_utility_prompt

Use only after a main workflow has already run.

Available utilities:

• clarity_rewrite
• conversion_rewrite
• tone_calibration
• structure_rebuild
• audience_rewrite
• impact_compressor
• blind_spot_finder
• shortcut_strategy
• ultra_leverage
• momentum_amplifier

generate_next_action

Use this when you already have output and need exactly one immediate next step.

Prompts

The server also exposes prompts for higher-level MCP clients.

Mode prompts:

• clarify_task
• strategize_task
• build_output
• improve_persuasion
• force_execution
• review_optimize

Domain prompts:

• run_freelancing_workflow
• run_products_workflow
• run_content_workflow
• run_execution_workflow
• run_investing_workflow
• run_utility_workflow

Front-door prompt:

• route_and_run

Resources

Full workflow resources:

• workflow://os/v4
• workflow://freelancing/v4
• workflow://products/v4
• workflow://content/v4
• workflow://execution/v4
• workflow://investing/v4
• workflow://utility/v4
• workflow://execute-referencing/v4

Section resources follow this template:

workflow://{domain}/v4/{sectionGroup}/{sectionName}

Examples:

• workflow://products/v4/execution/sequence
• workflow://content/v4/output/response-shape

Recommended Client Flow

For most client integrations, use this sequence:

1. Call route_task
2. Optionally read resource_uri or a section resource
3. Call run_workflow_sequence
4. If needed, call apply_utility_prompt once
5. Return the final output with the supplied next_action

Client Configuration

For client setup across Codex, Claude, ChatGPT, Kimi, Claw-style clients, and VS Code extensions, use the generic MCP configuration guide:

• Generic MCP Client Configuration

Notes

• The bundled workflow markdown files are the source of truth.
• The server is designed to run one main workflow per request.
• Utility prompts are supporting tools, not the primary entry point.
• Output should end with one immediate next action.