TCPFN - TEMPORAL CAUSAL PRIOR-DATA FITTED NETWORK
The Causal Engine Behind
Every
AI Analyst
TCPFN is the temporal causal foundation model that powers ProfitOps AI Analysts. It reads raw multivariate time series and recovers what actually causes what—so every recommendation is backed by causal reasoning, not guesswork.
90%
2 Min
Root cause identified
1 model
Replaces 3 separate tools
98%
Recommendations adoption
One Model · Three Jobs
From Sensor Data to Decisions
Most pipelines stitch together a discovery tool, a separate estimator, and a hand-built RCA script. TCPFN does all three with a single forward pass.
See What's Breaking
Recovers the cause-and-effect graph directly from multivariate time series — directed edges, time lags and all — without prior structural assumptions.
Find Root Cause
When something drifts, TCPFN traces the anomaly back through the graph to the upstream signal that started it — not just the correlated symptom.
Fix It Fast
Quantifies how much one variable actually moves another, so you can simulate “what if we change X?” before touching the plant.
Three Ways To Use It
Put TCPFN to Work, Your Way
One model, three modes — from a one-off investigation, to an always-on digital worker, to a tool inside your own agents.
Use case 1
Trace Any Anomaly, On Demand
Bring your time series and get a causal graph, root causes, and effect estimates in seconds—perfect for understanding what’s driving your business outcomes.
Use case 2
Powering a Digital Worker, AI Analyst
Give an AI Analyst an objective and TCPFN runs the loop 24/7 — an always-on digital worker that discovers the levers and holds your KPI on target.
Use case 3
Inside Your Agents, via MCP
Call causal discovery, root-cause analysis and effect estimation as native tools from any assistant or copilot through our MCP client.
01
Use case 1 · Trace Any Anomaly, On Demand
Watch TCPFN Recover a Causal Graph
Pick a benchmark dataset and watch TCPFN reconstruct its causal graph, root cause, and accuracy — live in your browser, no sign-up. To run it on your own data, create a free account.
02
Use case 2 · From model to teammate
This Is What Makes an AI Analyst's Plan Proven
Give a ProfitOps AI Analyst an objective and it doesn’t guess from correlations. TCPFN lets it reason over real cause and effect — turning a recommendation into an action plan you can trust.
Give it an Objective
Tell your AI Analyst the goal and the constraints.
“Hold the tank at 62.5 °C while cutting energy.”
POWERED BY TCPFN
Reason Over Cause & Effect
TCPFN discovers the causal levers, traces root causes and estimates the impact of each move.
Discovery · root-cause · effect estimation
Get a Proven Action Plan
The Analyst acts on the real driver and holds the KPI on target — and shows its work.
KPI back on target · +$600/hr realized
The digital worker, live ↓
TCPFN Doesn’t Just Analyze, It Can Run the Plant
Pearl is where ProfitOps AI Analysts go to work — an always-on digital workforce. Give an Analyst an objective and TCPFN closes the loop: discovering the causal levers, recommending actions, and holding the KPI on target 24/7. This is the real dashboard.
03
Use case 3 · Inside your agents (MCP)
Call TCPFN From Any Agent, Over MCP
TCPFN also ships as a Model Context Protocol server, so any assistant or copilot can use causal discovery, root-cause analysis and effect estimation as native tools — the same model behind the playground and Pearl.
01
Drop-in tools: discover_graph , find_root_cause , estimate_effect
02
Works with Claude and any MCP-compatible client
03
Point it at a live data source and ask in plain language
For Builders
Want to Build Your Own?
TCPFN was built with PFN Studio — our platform for building, training and fine-tuning prior-fitted foundation models. Bring your own priors and data, and ship a model like this one.