COUNTERFACTUAL INTELLIGENCE FOR INVESTMENT TEAMS

Test the portfolio decision before the market does.

Qelly freezes the current portfolio and assumptions, tests candidate actions across plausible and adverse futures, and preserves the evidence behind the comparison for human review.

Event condition → portfolio impact → robustness test. Probabilities enter only when their provenance and calibration support them.

Use Qelly's live synthetic Portfolio tools from an MCP-capable agent. Private portfolio connections, governed approvals, and continuous monitoring are the next authenticated layer.

Explore agent tools Try the agent-guided demo Inspect the immutable evidence case Discuss a design partnership

Synthetic and uncalibrated foundation. No client data. No trading. Human review required.

This page simulates the client experience in the browser. The separate public endpoint at /demo/mcp sends only bounded synthetic tool inputs to Qelly; it accepts no holdings and persists no decision case.

AGENT-NATIVE / LIVE SYNTHETIC

One live interface for the agent your team prefers.

The public synthetic Portfolio MCP endpoint is live. Your agent can call the same governed calculation and evidence boundary demonstrated here.

Live interface

Your agent

Available to an MCP-capable client such as ChatGPT, Claude, or a Kimi-compatible developer client.

  • State a premise
  • Request a comparison
  • Retrieve evidence
  • Draft a summary
LIVE MCP / LIVE

Live governed boundary

Qelly

  • VERIFIED Freeze inputs
  • VERIFIED Compute results
  • VERIFIED Attach provenance
  • VERIFIED Return evidence digests
LIVE / SYNTHETIC
Public, non-persisting synthetic MCP, local browser demo, and pinned evidence case
NEXT / PRIVATE
OAuth-protected holdings, portfolio APIs, approvals, and audit records
RESEARCH
Event-conditioned world models, calibration, and portfolio-specific training after private access controls
OUT OF SCOPE
Custody, trade execution, investment advice, self-approved recommendations
See tools, setup guides, and capability boundaries →
Conceptual visualization — not model output or evidence.

Verified product proof

One portfolio. Three branches. A decision that changes under pressure.

The portfolio artifact stays frozen while a Shear Test changes only the disclosed importance of the same three Scenario Branches.

Frozen inputs, bounded candidate actions, and content digests are the same properties an agent client must be able to rely on.

Pass A / frozen inputs

Baseline ranking

Verified
  1. 01 Hold
  2. 02 Raise cash
  3. 03 Defensive rebalance
Shear

Pass B / adverse weights

Adverse ranking

Displaced
  1. 01 Defensive rebalance
  2. 02 Raise cash
  3. 03 Hold
Engine case artifact / SHA-256
098a1729fbb961d0d5ff35d71f207d1a340bcb92966e254e469a37e1f89fc9cc
Baseline output / SHA-256
3dda2a8f083e920cac309a3722d88f36c3525ccadaefa77919ca2fe942c1dade
Adverse output / SHA-256
cfd2400bd0e699c33dc0c983a1b01792741bd50774365cefff11a421662b829f
Kernel / adapter
strategy-engine-v1 / finance-linear-v1
Risk policy / solver
finance-risk-v1 / deterministic-linear-evaluator
Source commit
32d4f52b8215b0f1a6932e6cf086bf8047675b79
Warnings
None
Inspect the complete evidence ledger →

Intended professional workflow

From a portfolio snapshot to a reviewable decision packet.

  1. 01 / ImportBring a governed portfolio snapshot.
  2. 02 / ValidateResolve coverage, constraints, and assumptions.
  3. 03 / Stress-testApply approved plausible and adverse branches.
  4. 04 / CompareRank the same candidate actions under the same frozen inputs.
  5. 05 / ApproveKeep a human reviewer in the decision path.
  6. 06 / ResolveReturn to what happened and improve the evidence.

Trust layer

Every conclusion keeps its assumptions attached.

Coverage before confidence

Unsupported exposures remain visible instead of disappearing into a score.

Immutable run identity

Inputs, versions, warnings, and content digests travel with the result.

Human approval

Qelly compares approved candidate actions; it does not execute or approve its own recommendation.

Two products, explicit boundaries

Professional portfolio assurance and event decisions remain separate.

Flagship / foundation

Qelly Portfolio

Pre-trade stress testing, scenario analysis, action comparison, and evidence retention for professional investment teams.

Separate application

Qelly Event Markets

Event-decision tooling and public calibration research with its own data, sessions, controls, and legal context.

Understand the boundary →

Measured research

New methods earn their place through measured comparisons.

Qelly starts with deterministic and classical baselines, then evaluates generative and quantum challengers through common contracts and end-to-end evidence.

Read the research approach →

DESIGN-PARTNER PROGRAM

Bring one recurring portfolio decision to a Qelly pilot.

We are looking for professional investment teams willing to define one governed decision, inspect the assumptions, and evaluate whether the resulting comparison is useful. Bring the agent surface your team prefers so we can define the right future handoff without implying it is connected today.

Write to luis@qelly.ai