Open-source · Self-hosted alpha

Easy to buy.
Hard to know when to sell.

Challenge every position. Cut what no longer works. Stay with what still does.

PortVise tests your investment thesis against live market data, recent news, portfolio risk, and changing market regimes, then turns the evidence into a structured decision memo.

MIT licensed OpenAI-compatible Docker ready
Local · Multi-agent review
Running
Pipeline
Plan Data Risk Regime Manager
Positions and conviction notes
6 positions
TickerWeight1YStatus
NVDA24.8%+41.2%Review
MSFT20.2%+18.6%Hold
GOOGL17.5%+12.4%Hold
AMD15.1%−8.7%Thesis risk
Cash8.0%Buffer
The decision gap

Conviction gets you in.
Evidence should decide what stays.

Selling is harder because the position is no longer an idea. It has a price, a history, and a story you have defended. PortVise gives every holding a fresh review.

01

The thesis changed

News, competition, or execution moved against the reason you originally bought.

Find the break
03

Volatility looks like failure

A drawdown can trigger an early exit even when the evidence still supports the position.

Protect the upside
Structured review

One portfolio. Eight focused agents.
A decision you can inspect.

PortVise separates research, risk, regime, and validation before the final recommendation, so the answer is built from visible evidence instead of one opaque prompt.

01

Planner

Frames the review and research priorities.

02

Data

Loads prices, returns, history, and macro signals.

03

News

Researches current events and thesis risks.

04

Risk

Measures factors, concentration, and stress losses.

05

Regime

Compares the market with historical analogs.

06

Theme

Finds hidden bets, crowding, and portfolio drift.

07

Validation

Checks the evidence for gaps and conflicts.

08

Manager

Turns the findings into prioritized actions.

Make every position earn its place.

The final memo explains the recommended posture, primary risk, timing, and what would justify changing the decision later.

  • Hold, reduce, hedge, rotate, add, or exit
  • Position-level rationale and sizing guidance
  • A clear do-nothing case
This week

Keep the thesis. Reduce the concentration.

The portfolio’s AI exposure remains supported, but correlated holdings create more downside than the individual theses suggest.

Posture
Selective trim
Primary risk
Factor concentration
Revisit
After earnings
UrgentReduceNVDATrim toward 18–20%
WatchHoldMSFTThesis remains intact

See the risk behind the tickers.

Estimate factor loadings, marginal risk, correlated clusters, and scenario losses.

Growth factorHigh
ConcentrationElevated
Rate sensitivityModerate

Ask how this setup behaved before.

Compare today’s macro state with historical periods and inspect forward portfolio outcomes.

Match78%Q4 2018
Avg return+4.8%21 days
Max drawdown−9.2%21 days
The PortVise Daily Brief

Three portfolios. One disciplined
end-of-day review.

Follow three transparent candidate portfolios as PortVise reviews the day’s evidence, challenges every position, and sends the decisions worth paying attention to.

01 Candidate portfolio

Compounders

Durable businesses with strong returns on capital and long reinvestment runways.

MSFTVCOSTGOOGL
Lens
Quality
Risk
Valuation
03 Candidate portfolio

Resilient Income

Cash-generative holdings selected for income, balance-sheet strength, and defense.

JPMPGJNJNEE
Lens
Income
Risk
Rates
PortVise Arena

Don’t just claim better decisions.
Put them in the arena.

Arena runs a paper-trading comparison between a plain LLM trader and the same trader informed by PortVise. Both start with the same capital, watchlist, rules, and market data.

01

Same starting lineIdentical portfolio and constraints.

02

Daily decisionsScheduled, recorded, and inspectable.

03

Honest evidenceAn experiment, not a promise of returns.

Arena
Active

advisor-arena

12 rounds · benchmark SPY
BaselinePlain LLM
$25,640 +2.56% total
Holdings
7
Max DD
−4.1%
PortVise-AdvisedLLM + PortVise
$25,910 +3.64% total
Holdings
6
Max DD
−3.2%
Equity curve
Baseline Advised
Illustrative interface data
Run it your way

Your portfolio. Your models. Your machine.

PortVise supports OpenAI-compatible endpoints, from cloud providers to local gateways. Run the web app with Python or package it with Docker.

MIT licensed OpenAI-compatible Cloud or local models CLI and MCP access
portvise / quickstartbash
# Install dependencies
$ uv sync
$ npm install

# Start the portfolio advisor
$ uv run python run.py

✓ PortVise running at http://localhost:7860

Decision support, not a crystal ball.

PortVise is a self-hosted alpha. Model output can be wrong, market data can be stale, and third-party providers may receive portfolio inputs depending on your configuration. It is not financial advice.

Read security notes
Review the evidence

Make the next portfolio decision
with more than a feeling.

Follow three portfolios and get their evidence-based review after every market day.