Customer success
About us

An agent built for data and finance tasks.

Ask any question of the filings. Every figure in the answer traces to the filing it came from, and every call the agent made is on record.

Get started for free Sign up
Ask, then trace
Compare, trace and open any statement from the chat. Click a figure and see where it is reported, what it is made of, and the filing line it came from.
No figure without a data call
The agent reads the filings through its tools, and only those. Every call behind every answer is kept; a figure no call returned is flagged.
Your team, your model
Invite your team, set who may read, run and rule. Bring your own model key; nothing else touches it.
Measured in the open
Run the public finance benchmarks against your own workspace and see every answer beside its reference and grade.

How it works

Sign up, add your model key, and ask. The agent reads the filings and shows its working.

STEP 1
Sign up
Your team gets its own workspace and agent, kept apart from every other team.
STEP 2
Add your key
An owner saves an Anthropic, OpenAI or DeepSeek key once. It is sealed on the server.
STEP 3
Ask
The agent reads the filings through its tools and opens comparisons, traces and statements in the chat.
STEP 4
Check it
Every figure opens its trace to the filing, and every data call behind the answer is on record.

Free for solo. Priced for teams.

Solo
Free
Your own agent over the filings, on your own model key.
Start free
Your own workspace and harness
Compare, trace and open any statement from the chat
Every figure traced to its filing, every data call on record
Bring your own Anthropic, OpenAI or DeepSeek key
Team
$63 /seat/mo · billed yearly
One workspace for the desk, with roles and a record of everything.
Start the team plan
Everything in Solo
Owners, analysts and viewers, with grants per project
One set of model keys for the team, sealed on the server
What every answer cost, by project and by member
Public finance benchmarks run against your workspace

What people do with it.

Compare a peer set

Ask for a line across companies and periods; the comparison opens in the chat with every figure linked to its filing.

Defend a number

Click any figure and see where it is reported, what it is made of, and the call the agent made to read it.

Check the agent

See every data call behind an answer, any figure no call returned, and how the workspace scores on public benchmarks.

Answers

What does the agent read?

The filings loaded into aggviz — their statements, facts and the links between them — through a fixed set of tools. It has no web access and no other tools.

Where do the numbers come from?

Every figure is a fact parsed from the filing's XBRL and carries its filing and fact key. The agent quotes facts; it does not produce figures of its own.

How do I know an answer is right?

Every data call behind an answer is kept and shown. A figure in the answer that no call returned is flagged, and any figure opens its trace back to the filing.

Which model does it use?

Yours. An owner saves an Anthropic, OpenAI or DeepSeek key for the team; it is sealed on the server and used only by the team's own workspace.

Who can see my team's work?

Only your team. Each team's workspace, sessions and keys are kept apart, and owners decide who may read, run and rule.

Writing.

All writing →
How we measure extraction recall — and why most pipelines can't
Jul 2026 · 6 min
The part of XBRL every vendor throws away
Jul 2026 · 6 min