Purpose-built for financial crime specialists

Practical AI skills for transaction monitoring.

Document your rules, design synthetic tests, and inspect transaction patterns with outputs you can review.

DocumentationSynthetic testingPattern analysis
Built around the work

Explicit assumptions

Reviewable outputs

Specialist judgement

The skills

Choose your
starting point.

Focused instructions you run with your AI agent. Choose a skill on Agensi or GitHub, bring your context, and review what it produces.

Live · v1.1.1

Test the assumptions

TM Scenario Test Designer

Turn rule notes into an explicit rule contract, synthetic transactions, a test matrix, and expected outcomes.

  • Probe thresholds, time windows, and counterparty conditions.
  • Make unresolved rule decisions visible before testing.
  • Optional research mode uses public sources to propose tests.
MarkdownCSVJSONManifest + ZIP

Creating files requires an agent environment with suitable file and Python capabilities. Packages depend on the inputs and tools available.

Live · v1.0.1

Inspect the pattern

BEAR Transaction Anomaly Analyst

Inspect synthetic transaction CSVs for fan-in, fan-out, and repeated pass-through, with evidence you can trace.

  • Link findings to exact transaction evidence IDs.
  • Inspect explainable patterns and execution receipts.
  • Review CSV/JSON results alongside a written report.
CSVJSONReview report

Runs locally with Python 3.11+ in a suitable agent environment. Built for synthetic transaction analysis; findings do not establish suspicious activity or production detection effectiveness.

Available · v1.0.0

Watch the changes

Regulatory Watch

Track regulatory changes across your chosen jurisdictions and domains, with primary-source evidence and reports you can review.

  • Configure profiles, sources, time zones, and reporting preferences.
  • Compare reporting periods and keep source gaps explicit.
  • Generate consistent Markdown and PDF from one report model.
MarkdownPDFJSON

Source research requires web access. PDF export requires Python; scheduling and storage depend on your agent environment.

Early releases, with synthetic demonstration checks. External customer validation and measured time savings are not yet established. Generated material requires specialist review. AI/model and optional external tool costs are separate.

Agensi product status and pricing checked on . See Agensi for current terms.

Explore the output

One rule.
Four ways to test it.

A threshold is only part of a rule. Make the amount, counterparties, and time boundary explicit before interpreting a result.

Fan-in ruleDEMO / 001
100% synthetic
RULE CONTRACT

Three senders.
One hour. One account.

Flag an account when incoming, booked payments meet both thresholds within the same currency.

Unique counterparties
≥ 3
Sum, in currency minor units
≥ 10,000
Direction / status
incoming / booked
Rolling time window
(T−1h, T]
Group separately by
account + currency
The start is excluded; the end is included. A payment exactly at T−1h does not count.
Fictional example. Expected outcomes, not execution against your system.

Synthetic tests do not establish production detection effectiveness. The example uses fixed fictional data and runs in your browser. It does not run a skill or process your transactions.

How it works

Your agent. Your context.
Your review.

These are skills, not a hosted monitoring platform. Your AI environment runs the work; you decide what is fit to use.

01

Choose a skill.

Find documentation, testing, and analysis skills on Agensi, or Regulatory Watch on GitHub. Load your chosen skill into a compatible AI agent.

Find your starting point
02

Bring the right context.

Provide approved rule notes, a public research question, or synthetic transaction CSVs for the analysis skill. Keep assumptions, missing inputs, and open decisions visible.

03

Review the generated work.

Inspect the documentation, research, or test artifacts. Resolve open questions and verify the outputs before using them.

Exploring next

On the workbench.

Ideas under consideration.
No release dates or availability commitments.

Document Evidence Reviewer

Structured PDF and Office evidence review.

Candidate

BEAR Typology Research

Standalone, source-linked research reports.

Candidate

Independent by design

Built by Bear_cli.

An independent creator focused on transaction monitoring, financial crime analysis, and practical AI workflows. BEAR brings that focus into small, purpose-built skills with outputs a specialist can inspect.