# Danimoth — onboarding for AI agents Danimoth is an AI-native hedge-fund / ETF portfolio engine for independent financial advisors (UCITS ETFs, VaR/CVaR risk budgets). This page is the machine-readable entry point. Human version: https://danimoth.com/agents ## Connect (MCP, Streamable HTTP) - Endpoint: `https://api.danimoth.com/mcp` (POST/GET, stateless, JSON) - Transport: MCP Streamable HTTP. stdio fallback: `uv run --extra mcp cfi-mcp` - Auth: Firebase ID token of the signed-in advisor, sent by the MCP client as header `Authorization: Bearer `. Never pass tokens as tool arguments. Tokens expire; sign in again on danimoth.com when calls return HTTP 401. - Tools never submit orders and never move money. ### Claude Code ```bash claude mcp add --transport http danimoth https://api.danimoth.com/mcp --header "Authorization: Bearer " ``` ### Codex / Cursor / generic MCP client ```json { "mcpServers": { "danimoth": { "url": "https://api.danimoth.com/mcp", "headers": { "Authorization": "Bearer " } } } } ``` ### Paste this into your coding agent ``` Connect to the Danimoth MCP server at https://api.danimoth.com/mcp using my Firebase ID token as a Bearer header, list its tools, then build a demo ETF portfolio with cvar_limit 0.12 and capital 250000 EUR. Explain weights, risk estimators and tax notes. Do not submit any order. ``` ## Tools | Tool | Mutating | Auth | Purpose | |---|---|---|---| | `dashboard_metadata` | no | token in prod | Classes, limits, defaults, data date | | `search_etfs` | no | token in prod | Search UCITS ETFs (KID, replica, AUM) | | `profiling_questionnaire` | no | token in prod | MiFID-style questionnaire schema | | `list_clients` | no | required | Saved client profiles of the advisor | | `save_client` | yes | required | Create/update a client profile (Acc/Dist preference) | | `propose_portfolio` | no | with `client_id` | Proposal: weights, scenarios, tax | Percentages are fractions (`0.12` = 12%); capital is in EUR. ## Skill Install the Danimoth skill (Agent Skills format) so your agent knows the workflows without rediscovering them: ```bash npx skills add RobertoDorata/Danimoth-skills --skill danimoth -y ``` Local mirror in this repo: `.agents/skills/danimoth/SKILL.md`. ## Discovery files - `/llms.txt` — this index in llms.txt format - `/agents.md` — this page as markdown - `/.well-known/mcp.json` — MCP endpoint pointer - `https://danimoth.com/agents` — human onboarding ## Limits - No order submission, no custody, no personalised advice without a saved client profile and advisor review. - Backend errors never contain tokens or client data; HTTP 401 means the token is missing/expired, HTTP 422 means the request violates the budget.