Headless GPX
The AI operating layer for private markets. Available to Claude, Microsoft Copilot, ChatGPT, and any MCP-compatible client.
Two surfaces. One operating layer.
Headless GPX is the open surface—bring your fund operating system into the AI of your choice. Inside the platform, JunieAI drives private-markets-native agents on the same data, permissions, and audit trail.
The Reality: As AI becomes more powerful, the operating layer matters more, not less
Yes, the models have tremendous raw power, but hand-rolled prototypes and generic chatbots running on exported spreadsheets won't survive contact with a real fund. They don't know your data model, can't see your permissions, and produce answers compliance can't sign off on.
AI needs an operating layer underneath it—a single source of truth for every team, every tool, every model.
The operating layer: 12 years of private markets operations, encoded
The data model, workflows, and trust framework that 2,300+ GPs rely on.
Data model
Funds, share classes, commitments, positions, capital calls: each is a deep concept in Juniper Square, not a field bolted onto a generic CRM. Our system knows a commitment isn't a contribution, so reports tie out, NAV rolls up, and AI gets the answer right.
Relationship graph
Every fund, asset, investor, prospect, and position is connected. Agents inherit the same context humans use, with LP to vehicle, GP to LP, and position to commitment relationships intact, and no extra mapping required.
Workflow library
Capital calls, subscriptions, AML/KYC, distributions, reporting cycles. The operational sequences that already run your fund, available as a tool surface designed for end-to-end work.
Permissions & identity
Agents inherit your existing Juniper Square configuration. The same row-level and document-level controls that govern people govern the AI working on your behalf.
Compliance & audit
Every read and write is logged against an identity, governed by your compliance posture, and exportable for review. The audit trail extends to the agent, not just the human-in-the-loop.
Controls gateway
A regulated control plane for the AI working across the fund. As employee-built, third-party, and agent-of-agent workflows proliferate, the system of record stays intact.
AI in your hands: Headless GPX, at work
How IR, finance, and compliance teams put Headless GPX to work.
More time on relationships
An analyst working in Claude or Copilot can prep faster and show up sharper for every LP touch point.
- Build a fully formatted investor tear sheet ahead of an LP meeting
- Pull recent positions, activity, and communications in seconds
- Surface talking points from an LP's portfolio history before a call
Less time on the close
CFO and compliance teams working in ChatGPT can accelerate the period-close while staying in control.
- Generate PCAPs and tick-and-tie financials at period close
- Flag investors with stale AML/KYC documentation
- Surface low cash balances ahead of distributions
We've been waiting for something like this. The ability to pull Juniper Square investor data directly into Claude or Copilot and have it just work, with all the context and controls already in place, is exactly how IR teams should be operating. Imagine quickly asking for an LP tear sheet before a meeting, right inside the AI tool I'm already using. That's the kind of workflow that could really change how IR teams operate.
Katie Riester
Head of Investor Relations
Felicis Ventures
Headless GPX is open. Bring it into your stack.
Connect any MCP-compatible client to your fund operating system. Your IR, finance, and compliance teams get the leverage of AI—you keep control of the system of record.
FAQs
-
At launch, our MCP tools cover read-access across contacts, accounts, offerings, prospects, investment entities, positions, subscriptions, documents, activities, communications, and more—with coverage continuing to expand. Our developer API runs alongside the MCP surface for system-to-system integrations. Write capabilities, agent-driven actions, and a headless LP Portal are on the near-term roadmap.
-
Headless GPX uses the Model Context Protocol (MCP), an open standard for connecting AI clients to tools and data sources. Any MCP-compatible client can connect to Juniper Square, including Claude, Microsoft Copilot, ChatGPT, Gemini, or a custom IDE.
-
Headless GPX inherits authentication, permissions, and audit from your existing Juniper Square configuration. Every agent operates within the same trust boundary as the humans on your team. You decide what data each agent can see and what actions it can take.
As employee-built agents, third-party agents, and agents-of-agents proliferate inside your firm, Juniper Square maintains the single regulated control plane for every action that touches the fund operating system. -
Models keep getting stronger, but the bottleneck for AI in the private markets is rarely the model. It's the layer underneath: fragmented data, inconsistent processes, permission structures that must be respected, legacy systems, compliance requirements, and workflows shaped as much by tacit knowledge as by formal policy. That's the operating layer. The model layer keeps moving. The operating layer is more durable.
-
Headless GPX is the open surface—it brings your fund operating system into whichever AI environment you choose. Inside the Juniper Square platform, JunieAI drives a family of private-markets-native agents on the same data, permissions, and audit trail. Customers get both best-of-breed AI built natively for private markets, and the freedom to use the AI environment that works best for their team.