Lets you carry a portable AI identity file (soul.md) across any MCP-aware client like Claude Desktop, Cursor, or Zenflow. The server exposes tools to read context, inject identity prompts, query skill depth, and write back decisions or lessons learned. Everything lives in a plaintext file at ~/soul.md that you version in git and own outright. Install is one npx line in your MCP config. Useful if you want consistent voice and context across different LLMs without vendor lock-in, or if you're building tools that need to read user identity as structured data rather than system prompts.
Your identity, not theirs. One file, every AI.
soul.md is an open format for your AI identity — values, voice, skills, taste. A plaintext file you own, version in git, and carry between tools. Manoma is the reference MCP server that makes any LLM read your soul.md.
See the soul.md format specification.
One line in your Claude Desktop, Cursor, or Zenflow config:
{
"mcpServers": {
"manoma": {
"command": "npx",
"args": ["-y", "manoma-mcp"]
}
}
}
Then drop your soul.md at ~/soul.md.
cp templates/founder.md ~/soul.md (also: engineer.md, designer.md, pm.md).Read — get_context(mode?), get_injection(mode?, message?), get_skill_depth(skill), list_sections().
Write-back — add_decision(...), add_lesson(...), update_now(...).
Resources expose every section as soul://section/<path>, plus soul://full for the whole file.
Full details in mcp/README.md and SPEC.md.
manoma/
├── mcp/ — MCP server (npx manoma-mcp)
├── templates/ — starter soul.md files
├── SPEC.md — format specification
└── README.md
Every AI vendor wants to own your memory. soul.md is the opposite bet: your identity as a plaintext file you control, interchangeable between tools, diff-able in git, outliving any individual model.
The schema is the thing. The runtime is just how it's read.
Contributions, forks, and parallel implementations welcome.
MIT.
SOUL_MD_PATHPath to your soul.md file. Defaults to ~/soul.md if not set.
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent