Data · Data Explorer

See what's in your data.Turn it into a policy.

Browse every document and table in a data source. Filter it, and the filter becomes the boundary your agents run inside.

Try it yourselfTalk to us
Free to start. No credit card.
SUPPLIER
Acme Coatings412
Harbor Paints288
Northline Chemical151
REGION
West530
Central207
East114
FILTER → POLICY
execution_policy
  supplier in [Acme Coatings,
               Harbor Paints]
  region = West
  → 418 documents in scope
Sample project · sample values
DOCUMENTSUnderstood, scored and sourcedDATAStored for every way it's askedAGENTSScored, inside Hardrails
THIS PAGE DELIVERSControlSee what an agent can reach before it runs.

Before it goes live

What can this agent see?
Answer before it runs.

Before an agent goes live, someone always asks what it can reach. Data Explorer lets you answer with the data itself, not with a prompt.

The Meibel console: a policy condition, project equals 450-columbus, and the list of exactly what the agent can read, 18 of 24 documents.12
  1. 1

    The condition you set.

    One line, on a field already in your data. No prompt to word carefully.

  2. 2

    Exactly what it can read.

    18 of 24 documents, listed by name, before the agent runs once.

Real console · sample project · public bid documents

What you see

Everything in a data source.
Before any agent sees it.

Every indexed field with its values and counts, the pieces your agents will search, and every table and column pulled from your documents.

Explore · supplier-safety-data · 948 documents
SUPPLIER
Acme Coatings412
Harbor Paints288
Northline Chemical151
TerraGrip97
REGION
West530
Central207
East114
Unassigned97
DOCUMENT TYPE
Safety data sheet604
Certificate221
Spec sheet98
Letter25
WHAT AN AGENT WILL FIND
search "flash point closed cup"
9. Physical properties. Flash point: 38 °C (closed cup).0.94Ironclad Primer 220 · p.2 · Acme Coatings
Flash point (Pensky-Martens closed cup): 142 °F.0.91Harbor Seal Coat · p.4 · Harbor Paints
Flash point: 23 °C, method not stated.0.78TerraGrip Epoxy · p.3 · scan
supplier_safety_data · HOW FULL EACH COLUMN IS
product_name100%
supplier100%
flash_point_c97%
hazard_class95%
batch_number88%
revision_date81%

SAMPLE PROJECT · SAMPLE VALUES

From filter to policy

The view you build
is the boundary it runs in.

Scope by what's actually in the data. The metadata you index is what you scope on, so there are no separate rules to keep in sync.

The Meibel console's Data Explorer: a filter on bid document type equals addendum, matching 6 documents.12
The same filter compiled into an execution policy, shown as JSON with a copy button.3
  1. 1

    Filter on a field you indexed.

    Supplier, region, document type. Here: addendum.

  2. 2

    See exactly what matches.

    6 documents, listed by name, before any agent runs.

  3. 3

    It compiles into a policy.

    Attach it to an agent. Hardrails enforces it on every search, query and tool call.

REAL CONSOLE · SAMPLE PROJECT · PUBLIC BID DOCUMENTS · HOW HARDRAILS WORK →

Browse your own documents as data.

Upload a folder, filter it, and give an agent only what the filter shows.

Try it yourselfTalk to us

Check before you scope

Spot it before an agent does.

Low-scored pages, empty columns and duplicate revisions show up in the data, not in an agent's answer three weeks later.

supplier-safety-data · EVERY DOCUMENT, BY QUALITY
Scored above your thresholdLow scoreMissing a fieldDuplicate revision
12Scans scored below 0.8All from one supplier. Re-request or route to review.
17Documents missing batch_numberListed as gaps, never guessed.
4Superseded revisionsExclude them from scope before the agent sees them.

SAMPLE PROJECT · SAMPLE VALUES

Answer the audit

Know what every agent could see.
Prove it later.

When someone asks what an agent could reach on a given day, show them the policy and the data it covered, not a prompt and a guess.

MAR 02Policy west_suppliers created from the Explore view
MAR 03Attached to the supplier safety agent, v3
MAR 03 TO 312,140 sessions, every one inside the policy
MAR 14One session asked for a Central supplier. Not in scope, nothing returned.
APR 08Auditor asks what the agent could reach on March 14
What could the supplier safety agent reach on March 14?
POLICYwest_suppliers · unchanged since Mar 02
DATA IN SCOPE418 documents · 2 suppliers · West only
TOOLSsearch, sql · no email, no web
THAT DAY71 sessions · 1 out-of-scope request, nothing returned
The audit log, in Hardrails →

ILLUSTRATION · SAMPLE PROJECT

How it connects

Between the data and the agent.

Explorer is where the data you built becomes the scope your agents run in.

DOCUMENTSYour fieldsDeep Transformation output lands in a data source.Deep Transformation → DATAA data sourceStored for every way it's asked.Retrieval-Ready Data →
THIS PAGEExploredSee it, check it, filter it.
CONTROLEnforcedThe metadata you index is what you scope on.Hardrails → AGENTSAgents on topSearch, query and graph, inside their scope.Agent Builder →
NOT A WAREHOUSEMeibel reads from your databases and systems and sits beside them. Nothing has to move.

At work across industries

Scope every agent by what's in the data.

Built for any document-heavy work. Here's what it does in a few industries.

CONSTRUCTION

One bid per session

One agent answers on any bid, scoped to a single project for each session.

Sample project · public documents
UTILITIES

One jurisdiction

Scope a tariff agent to one state's schedules, so it never quotes another's.

What you can scope by
HEALTHCARE PAYER

One plan year

Scope a benefits agent to the plan and year a member is in.

What you can scope by
LIFE SCIENCES

One regulator

Scope a label agent to one market's documents when the question is about that market.

What you can scope by
What will you build?Try it yourselfSee it by industry

Built on Meibel

Teams already ship on it.

Meibel is a foundational technology for our Generative AI practice. Our customers demand accountability and transparency from AI solutions, along with seamless integration into their existing operations and workflows. Meibel's platform delivers on all fronts.
Paul LiggettCEO, Cloud Eleven
Meibel’s Generative AI Platform allows FAST, along with our subordinate Supporting Effort, to expand the scope of US Army projects we are bidding on. We now deliver AI within our solutions that opens new opportunities, ensuring the military of the transparency, accountability and confidence they require.
Dan WrotenSenior Vice President, Public Sector FAST
Featured in

Read the case studies →

Questions

What is Data Explorer?

The Explore view of a data source in the Meibel console. It shows every indexed field with its values and counts, the chunks your agents search, and every extracted table and column.

How does a filter become a policy?

The filters and conditions you build compile into an execution policy. Attach the policy to an agent, and Hardrails enforces it on every search, query and tool call.

What can I filter on?

The metadata fields you index when documents come in: supplier, region, document type, or any field you define.

Can I write policies without the console?

Yes. Execution policies are also available through the API.

Start with one workload. Scale efficiently.

See what's in your data before an agent does.

Sign up free, connect a data source and open Explore. Then turn the view into a policy.

Try it yourselfTalk to us
PALETTE