Build a fund-data agent in 60 seconds.

50+ MCP tools covering the full fund lifecycle. Plug your Claude, Cursor, ChatGPT, or custom agent into live deal, LP, capital-call, audit, and compliance data — same metering humans pay, full audit trail, human-approval gates on every write. Access is scoped in your firm's enterprise agreement.

MCP 2024-11-05 50+ tools published Same price as humans Try-before-you-pay Full audit trail

Talk to us about agent access → View tool catalogue

60-second quickstart — Claude Code

# 1. Add VantedgeAI to Claude (one line, OAuth flow opens in browser)
claude mcp add fundos https://fundos.vantedgeai.com/mcp

# 2. Ask Claude anything about your fund
> Find the LP with the largest commitment who hasn't been
  contacted in 60 days, and draft a follow-up.

# Claude calls (under the hood):
#   fundos_list_lps  →  filter by commitment + last_contact
#   fundos_get_lp(lp_id)  →  pull full record
#   fundos_run_agent("lp_fundraising_autopilot", lp_id)  →  draft
# Returns the drafted email for your approval before sending.

For other MCP clients

{
  "mcpServers": {
    "fundos": {
      "url": "https://fundos.vantedgeai.com/mcp",
      "auth": { "type": "bearer", "token": "vdr_YOUR_API_KEY" }
    }
  }
}

OpenAI Assistants / function calling

# 1. Fetch function-call declarations from VantedgeAI
curl https://fundos.vantedgeai.com/api/v1/tools/gemini-declarations \
     -H "Authorization: Bearer vdr_YOUR_API_KEY" \
     > fundos_functions.json

# 2. Pass them to your OpenAI Assistants thread
from openai import OpenAI
client = OpenAI()
assistant = client.beta.assistants.create(
  name="Fund Ops Agent",
  model="gpt-4o",
  tools=json.load(open("fundos_functions.json")),
)

Direct REST (for any framework)

# Tools are also accessible via plain REST — no MCP client needed
curl -X POST https://fundos.vantedgeai.com/api/v1/tools/fundos_list_deals/call \
  -H "Authorization: Bearer vdr_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"args": {"stage": "diligence"}}'

# Returns: { "deals": [...], "credits_charged": 1 }

Why agents pick VantedgeAI

Same price as humans

Tool prices are identical whether a human clicks the button or your agent invokes the MCP. No agent-tax, no different SKUs. View pricing.

Try before you pay

Packaged tasks (audit pack, ODD, CIM) charge only when the customer downloads, accepts, or chains the output. Refuse it or let 7 days pass → no charge.

Full audit trail

Every tool call writes a CreditLedger row with trace_id + duration + result. Customers see exactly what your agent did in their /admin/usage-traces dashboard.

Human approval gate

Mutating tools (create_*, send_*) write to an approval queue. The GP clicks Approve before anything leaves the building. Reduces customer fear of agent autonomy.

Trust signals published

30-day calls / success-rate / p95 latency per tool — see tools.json. Pick tools with track record, not just specs.

Bring-your-own-data

Tools can read from the customer's Google Drive, SharePoint, Egnyte, or any MCP source. No data migration; no copy in VantedgeAI's storage.

What you can build

Internal fund agent (live inside the GP)

Daily "what changed since yesterday" digest. Auto-screen incoming pitch decks. Draft LP follow-ups based on recent meetings. Run risk scans every Monday morning. Generate quarterly LP letters from portfolio data. Best for: a GP team building their own operational AI co-pilot.

Embedded agent in a fund-tech product

You're building DealCloud / Affinity / Carta / Brex-equivalent — and you want AI features grounded in customer data. Embed VantedgeAI's MCP as your data layer. Customer pays VantedgeAI per call; you keep the product revenue. Best for: vertical SaaS adding AI without rebuilding fund-data infrastructure.

Public agent product (Claude / ChatGPT marketplace)

Build a "fund analyst agent" anyone can install in Claude. They connect their VantedgeAI account; your agent does the work. Revenue flows through VantedgeAI's billing — you focus on the UX. Best for: agent-first startups looking for distribution.

Pricing for agent builders

Same tiered pricing humans pay. Pull /.well-known/tools.json programmatically to predict cost (and now reliability) before invoking:

Tier	Credits	USD	Examples
read_fast	1	$0.01	list_deals, get_lp
read_multi	2	$0.02	get_agent_context
read_compute	5	$0.05	run_pricer, compute_waterfall
write	10	$0.10	create_deal (human-approved)
ai_light	25	$0.25	vdr_analyze, generate_odd
ai_heavy	50	$0.50	generate_cim

Packaged tasks (audit pack, ODD, CIM, K-1, valuation memo) carry higher per-call prices reflecting outcome value, not LLM cost — sized into your firm's enterprise agreement. Talk to us to scope agent access.

Safety guarantees

No anonymous calls. Every request authenticates as a real VantedgeAI user with org-scoped data access.
Mutating tools queue for human approval. Your agent proposes; a GP clicks Approve. No surprise capital calls.
Customer data never copied. When agents read from BYOD sources (Drive, SharePoint, Egnyte), the docs stay in the customer's stack.
Rate-limited per-user, per-feature. A runaway agent can't bankrupt the customer in a single hour.

Discovery files

/.well-known/mcp.json — MCP server descriptor + tool count
/.well-known/agents.json — VantedgeAI's 9 built-in agents (Agent A–I)
/.well-known/tools.json — 50+ tools + prices + trust signals
/.well-known/oauth-protected-resource — RFC 9728
/llms-full.txt — full integration reference

Building something we should know about? Tell us — we ship new tools weekly and prioritise based on agent demand.