Powering the frontier of equity research.

State-of-the-art agentic AI infrastructure for equity research, built by analysts and AI researchers.

What we believe

Leading equity research teams are taking AI in-house.

A system you own is one you control, keep current and can push as far as the technology goes. Primer builds the specialist infrastructure those systems need.

Why funds choose Primer
  1. Workflow apps aren’t the end state.

    The frontier moves every month, and a workflow app locks in today’s process. Over time, research teams will want systems flexible enough to evolve with it.

  2. Agents need specialist infrastructure.

    Financial data, modelling tools, analyst-written methods, company memory and evals are what turn a general agent into an equity research system.

  3. Your system should keep getting better.

    Your system should improve with every new model and method. Built on state-of-the-art tools, it does, so your team’s job is the research process, not the plumbing.

  4. Faster research is only the start.

    Most AI in research today does the existing job faster. Investment processes will evolve as AI discovers new methods, and exploring those takes infrastructure that can test ideas at scale against real outcomes.

  5. Every team is moving as fast as it can.

    Whether you’re starting your AI journey or building your own system, Primer supports every stage. The market is moving too fast to be stuck on an app that isn’t.

Our products

Two ways to use Primer.

Both run on the same research tools, methods and evals. The key difference is where it runs: with us, or in your own environment.

The Primer web app

Ready today, with nothing to install.

  • Full research tasks, from filings to model to written note
  • Monitors your coverage and flags what matters
  • Works in the browser, or inside Claude and Copilot

Built forAnalysts and PMs who want state-of-the-art AI today

How the web app works

Forecasting track record

Forecasting is the ultimate eval for equity research.

A forecast tests the whole equity research process: understanding the company, building the model, forecasting, and the reasoning behind it all. Then it’s measured against an outcome that is economically valuable and unknowable in advance: what the company actually reports.

See the live record on OpenStocks
72.2%of Primer’s forecasts were closer to reported earnings than consensus

What every forecast tests

The data
The right numbers, from the right filings
The model
A model that turns them into a forecast
The call
Where to disagree with consensus, and why
The result
The company reports. The score is objective
Primer vs Wall StreetPrimer 211Wall Street consensus 105
050100150200250

Benchmarks

Proven on real analyst work.

Forecasting tests the process end to end. These benchmarks test the parts that drive it: retrieval and modelling first, then multi-step research questions on BigFinanceBench and our cash-flow eval.

See all the evals

FinRetrieval

Data retrieval

Primer0
Claude Opus 4.5 + Daloopa0.0
Gemini 3 Pro + Daloopa0.0
GPT-5.2 + Daloopa0.0
ChatGPT 5.2 (Web)0.0

Wall Street Prep

Financial modelling

Primer0
Shortcut.ai0
Claude in Excel0
ChatGPT0
Microsoft Copilot0

Primer eval

Underlying cash flow

Primer0.0
Fable 5 Max0.0
GPT-5.5 Ext. Thinking0.0
Opus 4.8 Max0.0
SuperGrok Heavy0.0
Gemini 3.5 Thinking0.0
“This is the best thing we've used by some margin. We love it.”
Analyst, single-manager hedge fund

Team

Domain expertise and deep AI capability, in one team.

Pushing the frontier of AI in equity research takes more than a strong technical team. It takes deep domain expertise and applied AI, together.

Why funds choose Primer
  • Seb Jory

    Founder
    Portfolio Manager

  • Ruggero Gargiulo

    Co-founder
    Ex-analyst and quant

  • Alistair Smallwood

    Head of Applied AI
    Ex-analyst, AI engineer

  • Chris Mingard

    CTO
    ML PhD, Oxford

Latest from our R&D team · Working paper · 28 Sep 2026

Systematic Alternative Data Research with Language Models

A research system that finds candidate datasets, reconstructs what was knowable at the time, and tests each relationship the way an analyst would. We publish methods like this, then ship them as tools.

Read the paper

In your control

Your agents. Your data. Your control.

Use Primer in the browser or inside the systems your team already uses. Your data is ring-fenced, never used to train models, and stays with your firm.

SOC 2 compliant
Security controls and materials ready for institutional review.
Web app or your own agents
Use Primer in the browser, or run its tools inside Claude Code, Codex or Cowork.
Sources included
Full filings for ~14,000 companies going back 10+ years, plus consensus and market data, with provenance kept.
Your data stays yours
Client data and customisations are ring-fenced and never used to train models.