Build Your Agentic Investing Team On GPAgent
GPAgent is the platform for GPs who want agents to operate like real investing team members.
GPAgent runs a hosted agentic investing team for fund managers. The platform owns integrations, orchestration, permissions, UI, guardrails, and the activity ledger. OpenClaw, Claude Code, Hermes, MCP clients, and custom agent fleets can connect through GPAgent's API and MCP surfaces, but GPAgent is the operating platform for the team. The goal is not automation theater. The goal is a sharper fund: better sourcing, better selection, better access, better portfolio support, and better LP outcomes.
This guide is generated from src/lib/agent-fleet.ts. Do not edit this file by hand. Run pnpm docs:agent-team after changing agent roles, and pnpm docs:agent-team:check before shipping.
The Investing Team
Sourcing Analyst
Finds alpha before it is obvious.
The Sourcing Analyst discovers companies from configured sources, enriches them with market and founder context, scores them against the GP’s investing lens, and introduces the most relevant companies into the pipeline with a clear rationale.
Investment Analyst
Turns sourced companies into investment attention.
The Investment Analyst reviews the best opportunities, writes reasoning a GP can trust, and moves the strongest companies into the human queue for founder conversations. It keeps the GP’s attention scarce and intentional.
Relationship Manager
Builds access.
The Relationship Manager acts like an EA for the investing workflow: calendar-to-CRM updates, founder connection queues, meeting context, follow-up actions, and card hygiene. It keeps source links, websites, deal notes, call recordings, and founder links attached to the right Deal Card when confidence is high.
Portfolio Analyst
Compounds alpha after investment.
The Portfolio Analyst reads portfolio updates, extracts metrics and risks, identifies where the GP can help, and turns company progress into structured intelligence for the GP and LPs.
GP Brand Strategist
Narrates alpha.
The future GP Brand Strategist will turn actual fund activity into public narrative: founder lessons, market themes, and public posts. It should be grounded in real GPAgent activity, not generic marketing prompts.
What GPAgent Owns
GPAgent is the durable center of the team:
- The fund's canonical company, deal, founder, relationship, portfolio, and LP data.
- API rails and scripts that agents use to write structured work.
- UI surfaces where a GP can inspect, approve, and act.
- Guardrails for tenant scoping, permissions, duplicate checks, ownership, and stage hygiene.
- Hosted orchestration for production agent runs.
- A RunActivity ledger that records what each team member did and whether the work actually succeeded.
GPAgent is the system of record and execution surface. External agent harnesses can connect through API and MCP without replacing the GPAgent-run team.
How The Team Works
The active team today is Sourcing Analyst, Investment Analyst, Relationship Manager, Portfolio Analyst. The Sourcing Analyst finds and qualifies companies. The Investment Analyst studies the best ones and prepares them for founder conversations. The Relationship Manager builds access and keeps relationship state current. The Portfolio Analyst supports companies after investment. The future GP Brand Strategist turns verified fund work into public narrative.
That sequence matters because each agent is doing a job a GP already understands. The product is not a bag of prompts. It is an investing team with clear ownership.
Why The Ledger Matters
Agentic work has to be auditable. A scheduler saying "ok" is not enough.
GPAgent's RunActivity ledger records who acted, what ran, what changed, what was checked, whether delivery succeeded, what failed, and what is safe to do next. That makes an investing team inspectable by a GP, recoverable by an operator, and safe enough to run across multiple funds.
A Simple Example
A GP connects fund data sources and enables a hosted sourcing run. GPAgent syncs configured company sources, ingests private source digests when available, enriches companies, writes only qualified deals, filters the rest with evidence, and records what happened. OpenClaw or another agent fleet can also call the same GPAgent APIs and MCP tools for custom workflows. The Investment Analyst reviews the strongest companies and moves a few into the GP's founder-conversation queue. The Relationship Manager turns qualified companies into passive founder-access prep, syncs real calendar activity back to the CRM, and attaches relevant meeting context. When the fund invests, the Portfolio Analyst tracks updates and turns progress into LP-ready intelligence.
The agents do real work. GPAgent keeps the work durable.
Current Roadmap
The current build order has completed canonical docs, single-writer cleanup, RunActivity implementation, Source Digest intake, and the first Relationship Manager operating loop. Next comes deeper integrations, richer person-level artifacts, more automated evidence gathering across the configured workspace, natural-run reliability, and Portfolio Analyst alignment.
Outreach prep is downstream of qualification. Deal Cards should carry the supporting context a GP would expect from a strong EA: source pages, websites, deal notes, call recordings, and founder links. Relationship actions should remain passive and internal until an operator explicitly enables external connectors. Public examples stay generic; customer-specific fund data stays inside authenticated workspaces.
Non-Divergence Rule
This public guide is generated from the canonical agent fleet config, not maintained as a separate source of truth. Public language translates the internal model into customer-safe product copy, but it must not create a different role model, roadmap, ownership boundary, or customer-specific example.