About Kela Fundos
Kela Fundos is an MCP server for FundOS fund operations, giving AI agents access to tools and data for deals, investor relations, compliance, and more. It also offers a document API that does not require a FundOS account.
Key features
- 01MCP tools for fund operations
- 02Deal and investor relations management
- 03Risk, compliance, and operations
- 04Document API without an account
- 05Seven-skill QM pack
Use cases
- 01Connect agents to fund operations data
- 02Generate K-1 and ODD/DDQ documents
- 03Analyze documents for risks
- 04Handle fund workflows through QM skills
MCP support
The remote MCP server exposes fund-operations tools to MCP clients, covering deal rooms, deal CRM, LP CRM, risk, pricing, CFO, OMS, compliance, and other areas. It connects over SSE or Streamable HTTP and uses OAuth 2.0 PKCE or Bearer API-key authentication.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://www.kela.com/mcpclaude mcp add --transport http fundos https://www.kela.com/mcp{"mcpServers":{"fundos":{"type":"http","url":"https://www.kela.com/mcp"}}}Open in Cursor{"servers":{"fundos":{"type":"http","url":"https://www.kela.com/mcp"}}}Open in VS Code{"mcpServers":{"fundos":{"type":"http","url":"https://www.kela.com/mcp"}}}{"mcpServers":{"fundos":{"serverUrl":"https://www.kela.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEfundos-mcp
MCP server for FundOS — connect Claude Code, Cursor, Codex CLI, and any AI agent to your fund operations.
What is FundOS?
FundOS is an AI-native operating system for fund managers — private equity, private credit, hedge funds, and venture capital. It replaces disconnected tools (Allvue, Juniper Square, Datasite, Salesforce) with a unified agentic platform across the full fund lifecycle.
Two ways to integrate
Option A — MCP server (full FundOS account)
Connect Claude Code, Codex CLI, or any MCP client to your live FundOS data.
Claude Code (SSE):
claude mcp add --transport sse fundos https://www.kela.com/mcp/sse
Codex CLI (Streamable HTTP):
# Step 1 — add the MCP server
codex mcp add fundos --url https://www.kela.com/mcp \
--bearer-token-env-var FUNDOS_API_KEY
# Step 2 — get a vdr_ key at https://kela.com/admin/api-keys
# Add to ~/.zshenv (not ~/.zshrc — Codex uses non-interactive shells)
export FUNDOS_API_KEY=vdr_your_key_here
# Step 3 — run (--sandbox danger-full-access is required for MCP tool calls)
codex exec --skip-git-repo-check \
--sandbox danger-full-access \
"Use the fundos MCP server to check my deals"
Codex CLI uses Bearer API key auth. Claude Code uses OAuth 2.0 PKCE — it opens a browser window on first use.
47 tools across: deal rooms, deal CRM, LP CRM, risk, pricer, CFO, OMS, syndication, HF Ops (DTCC), compliance, and more.
Option C — QM skill pack (YC's agent harness)
QM is Y Combinator's open-source multiplayer agent
harness. It cannot use the MCP server above — QM uses MCP as its own internal
in-process tool transport and seals it against external servers, so pointing it at
/mcp does nothing.
Instead, FundOS ships a seven-skill pack for QM that calls the REST API:
integrations/qm/. Skills cover capital calls, distribution
waterfalls, NAV and fund metrics, LP communications, covenant monitoring, and the data
room. Credentials go through QM's credential broker, so the agent never holds the key.
Option B — Stateless Document API (no FundOS account required)
Law firms, fund admins, LPs, and placement agents can generate K-1s, ODD/DDQs, and document analyses without creating a FundOS account or loading any fund data.
Get a doc_ key at https://kela.com/developers/register — email + password, key appears immediately. First 30 days free.
curl -X POST https://kela.com/api/v1/documents/k1 \
-H "Authorization: Bearer doc_your_key" \
-H "Content-Type: application/json" \
-d '{
"fund_name": "Eight Capital Fund II",
"tax_year": 2025,
"partner_name": "Aksia LP",
"lp_share_pct": 0.15,
"box_1_ordinary_income": 45000,
"box_9a_lt_gain": 105000
}'
# → { "html": "<!DOCTYPE html>...", "filename": "k1_2025_aksia_lp.html",
# "charged_usd": 5.00 }
Five endpoints — no account needed:
| Endpoint | Purpose | Price |
|---|---|---|
POST /api/v1/documents/k1 | Schedule K-1 (Form 1065) | $5/call |
POST /api/v1/documents/odd | ODD/DDQ completion | $25/call |
POST /api/v1/documents/analyze | Document red-flag + entity-map | $3/call |
POST /api/v1/documents/waterfall | European waterfall | Free |
POST /api/v1/documents/pricer | IRR/MOIC/WAL pricer | Free |
Quick links
| 📚 Docs | docs/quickstart.md |
| 🔑 Stateless API | docs/stateless-api.md |
| 🛠 MCP Tools | docs/mcp-tools.md |
| 🔒 Auth | docs/authentication.md |
| 🐍 Python example | examples/python-client.py |
| 🖥 Cursor setup | examples/cursor-setup.md |
| 🌐 API reference | https://kela.com/api/docs |
| 💬 Support | [email protected] |
Discovery files
https://kela.com/CLAUDE.md — canonical agent guide
https://kela.com/llms.txt — short FundOS overview for LLMs
https://kela.com/llms-full.txt — full developer + integration reference
https://kela.com/.well-known/mcp.json — MCP server discovery
https://kela.com/.well-known/oauth-authorization-server
https://kela.com/.well-known/oauth-protected-resource
Pricing (MCP / full account)
- $299/paying-seat/month includes 6,000 credits
- 1 credit = $0.01 — same for humans and agents
- Tool tiers: read_fast 1cr → ai_heavy 50cr
- Outcome fees (waived during 30-day trial): K-1 $50, ODD $150, CIM $250, audit pack $100
License
MIT
AX Benchmark
benchmarks/ax/ contains the open-source FundOS Agent Experience benchmark —
20 fund-ops tasks run under two conditions (curated FundOS docs vs a raw API
reference) and scored on correct tool choice, hallucinated endpoints, and
human-approval gate coverage. Live results: https://kela.com/ax
pip install pyyaml
export GEMINI_API_KEY=AI... # or ANTHROPIC_API_KEY (pip install anthropic)
python benchmarks/ax/runner.py
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.