The trading layer forAI agents and your ideas

Talk markets with Aura. Build baskets, test ideas and set the rules your agent trades by. Designed for Robinhood.

Try the demo

How Seer works

Recents

Is NVDA too concentrated? Show me a smaller allocation before changing anything.

Working1s

Explore the system

Markets move fast. Give your agent the whole picture.

Your agent works within your rules

Explore research, forecasts and portfolio proposals in one paper-trading workspace.

Size the idea for the whole portfolio

A constrained optimizer weighs return, correlation, liquidity and trading costs. Every proposed allocation stays inside your mandate.

Follow the event back to its source

Language models turn filings and news into structured events. Sources, timestamps, new information and contradictions stay attached to the thesis.

Build a basket around an idea

Discuss a market theme with Aura. Compare a focused basket, a broader allocation and a cash reserve, with limits and trade-offs visible before approval.

Give every decision a receipt

Link the evidence and data cutoff to the forecast, strategy version, policy result and eventual outcome. Keep assumptions separate from observations.

Track the order through every state

A separate execution service submits approved orders and reconciles the result. Proposed, submitted and filled always mean different things.

Put your limits into the system

An independent policy service checks eligibility, exposure, liquidity, data freshness and costs. Fixed rules validate every proposed order.

You: Review my concentration and propose a smaller NVDA allocation. Show me the changes before placing a paper order.

Aura

Reviewed 9 sample positions against your mandate. The rebalance is a proposal; approval is still required.

You: Explain what this market news could mean for my basket. Separate the evidence from the assumptions.

Aura

Source review is ready. Six event records are linked to the evidence. The downside scenario is still being evaluated.

Research drafts and their status
Asset	Status
Event records, 6 sources	Draft, ready for review
Scenarios, 3 horizons	Draft, in your queue
Newsletter	Writing...

You: Compare the possible outcomes with our Forecast Service. Show the assumptions too.

Aura

Seven features across three scenarios. These numerical examples are illustrative, not live forecasts.

Signal 1 of 7

A forecast is a distribution of outcomes.

This example compares expected return, downside and estimated trading costs. Before we...

Aura

I prepared three sample baskets around your idea. Each includes a defined universe, allocation limits and a cash reserve. All three are ready for review.

“a diversified technology basket”

Paper scenario
Sample idea
3 scenarios compared

Core Basket.

Basket

Core Basket. Start with a wider view.

Aura compares the thesis with numerical forecasts, estimated costs and your portfolio limits.

Core Basket.

Basket

Core Basket. Start with a wider view.

Aura compares the thesis with numerical forecasts, estimated costs and your portfolio limits.

You: Trace this proposal back to the evidence and ask the Policy Service to prepare a review.

Aura

Six checks prepared by the policy service. The proposal needs approval before the execution stage.

Review 1 of 6

To: research@seer.example
Subject: Decision receipt: evidence, limits and costs

Hi team,

I reviewed the example proposal against the mandate. The data cutoff, source references and strategy version are recorded. Numerical forecasts describe a range of possible outcomes rather than a guaranteed price. The optimizer evaluates the change across the whole portfolio, including concentration, liquidity and estimated costs. The policy service checks the proposed orders against refreshed execution data. This sample remains at proposed until the required approval is recorded. If the quote expires, the proposal must be refreshed. Submission and a confirmed fill are separate states, and an uncertain response must be reconciled before any retry. All holdings, forecasts, costs and activity here are illustrative. No real account is connected. The intended system keeps credentials outside language-model prompts and treats research text as evidence, never as order authority. Thanks, Seer Research

You: Prepare the paper order from this approved proposal. Show the checks, status and eventual fill.

Paper Execution Plan

Seer Brief

Order review thread

aura-order-review

Aura
APP
Paper order is ready for review. The mandate, quote and policy checks are attached.
PDF attachment: aura-order-receipt
Seer Research
The checks are clear. Keep the order at proposed until the paper approval is confirmed.
PDF attachment: aura-order-receipt

Aura

I checked the sample portfolio against your mandate. Concentration, liquidity and approval limits are recorded. The proposed rule changes need your confirmation.

Mandate checks by category
Location	Count
Company	11 instances
Journal	9 instances
About Us	7 instances
Press	3 instances
Homepage	1 instance

Your mandate is the boundary

You choose the capital, eligible assets, position limits and approval mode. Aura works within those rules.

Discuss the world with your agent

Explore an idea in plain language, then follow the sources and assumptions behind it.

Explore Market Signals

Inspect the forecast and the costs

Compare a range of outcomes over a defined horizon before deciding whether to act.

Explore Strategy Insights

Aura brings sample market signals and basket mandates together

Choose how much the agent may do

Approval, hybrid or autonomous within the mandate. New agents begin with approval in this preview.

Explore Your Rules

Intelligence with a job to do

Language models interpret. Numerical models forecast. The optimizer allocates. Fixed rules validate. A separate service executes.

01
Language models extract structured events from timestamped filings, news and other supported sources.
Aura
·
Event interpretation
Explore layer
02
Numerical models estimate a range of returns and downside over a fixed horizon.
Atlas
·
Numerical forecasting
Explore layer
03
A constrained optimizer balances expected return, correlation, liquidity and trading costs.
Pulse
·
Portfolio optimization
Explore layer
04
Independent rules check every proposed order against the mandate and refreshed market data.
Hedra
·
Deterministic policy
Explore layer
05
An isolated executor submits approved orders, resolves uncertain states and reconciles every outcome.
Aura Core
·
Separate execution

Every action leaves a record

Your Aura links the evidence to the forecast, checks and reconciled outcome.

Inspect a decision from end to end
Follow the data cutoff, strategy version, approvals and sample fill in one receipt.
Explore Agent Activity
Measure performance after costs
Compare returns, drawdown and turnover against a benchmark, including trading and research costs.
Explore Evaluation
Check the assumptions behind a thesis
Trace each proposed strategy back to its evidence and open questions.
Explore Source Checks
Keep forecasts open to new evidence
Monitor calibration and changing conditions without rewriting the original decision record.
Explore Market Sentiment
Keep identity separate from authority
Agent profiles, strategy versions and account mandates are separate records with explicit permissions.
Explore Your Rules

Built between intent and execution

Research ahead. A shorter order path.

Research and optimization run in the background. Refreshed quotes and fixed checks stay on the order path. Activity below is illustrative.

Open Command Center

Paper Runs

263,867

300K200K100K0

Apr 17Sep 7

All time

Example Activity by Category

Research
P99
Allocation Review
P47
Strategy Research
P91
Reporting & Monitoring
P75
Automation
P63
Order Reconciliation
P93

0P20P40P60P80P100

Test Sessions

525

6004002000

Apr 17Sep 7

All time

Agent Directory

2
Atlas / Research
3,140 Signal checks
1,642 Paper runs
1
Aura / Coordinator
4,039 Signal checks
1,728 Paper runs
3
Hedra / Policy
2,684 Signal checks
894 Paper runs

User		Signal checks	Paper runs	Test days
4	Pulse / Signals	1,811	594	136
5	Aura / Core Basket	1,177	403	141
6	Atlas / Index Basket	1,041	355	339
7	Hedra / Risk Review	959	440	124

Keep account details out of research

The planned system separates public research from account details and keeps signing credentials in the execution service.

Explore Privacy Controls

Planned Robinhood integration

Robinhood Trading MCP is the planned brokerage connector. Stock Tokens on Robinhood Chain require a separate verified execution route.

Explore Integration Scope

Understand the system behind the conversation

Research notes on forecasts, portfolio controls, execution and privacy. Explore the intended design behind this paper-trading preview.

See All

What a forecast range tells you
What a forecast range tells you (opens in a new tab)
By Seer Research · 1 Jul, 2026
The rules behinda consideredrebalance
The rules behind a rebalance (opens in a new tab)
By Seer Research · 13 Aug, 2026
When should your basket rebalance?
When should your basket rebalance? (opens in a new tab)
By Seer Research · 8 Jul, 2026

Who sets an agent’s limits?
Who sets an agent’s limits? (opens in a new tab)
By Seer Research · 16 Jul, 2026
Where do market signalscome from?
Where do market signals come from? (opens in a new tab)
By Seer Research · 30 Jul, 2026
When the edge is small:Why an agentcan chooseto wait
Why an agent can choose to wait (opens in a new tab)
By Seer Research · 25 Jun, 2026

From headlineto event record
From headline to event record (opens in a new tab)
By Seer Research · 9 Jul, 2026
Inside an order receipt:From proposal to outcome
Inside an order receipt (opens in a new tab)
By Seer Research · 9 Jul, 2026
An agent’s researchbudget: how computationstays separate frominvested capital
An agent’s research budget (opens in a new tab)
By Seer Research · 24 Jun, 2026