01
Least necessary access
During a pilot, connect only the data, models and tools a single clear use case needs, rather than exposing unrelated systems to the agent workflow at once.
Trust
FiClaw trust & security
A financial agent must state not only its capability but also its data, model, permission and human-review boundaries. This page covers FiClaw's basic principles for public trial and team pilots.
Principles
Basic principles for financial settings
02
Human-in-the-loop review
FiClaw assists research, analysis, review and pre-execution checks; key judgments should still be confirmed by team members against internal policy.
03
Boundaries before automation
Any workflow touching trading, risk control, client assets or sensitive internal data must first define permissions, logging, rollback and human-confirmation boundaries.
Data Boundary
Data usage boundaries
Website lead data
The site's Contact / Demo forms are only for inquiry handling, demo scheduling and partnership talks; the field scope follows the current form and privacy policy.
FiClaw Desktop usage data
Desktop login, model calls, membership and credit records support account, model routing and consumption display; the exact scope follows the in-product notes.
Institutional pilot data
A team pilot first confirms data sources, access method, usage scope, retention and responsibility split before entering real-workflow validation.
Model context data
Context sent to cloud models should be organized on a task-minimal basis; do not put unrelated sensitive material into a one-off request.
Controls
Controls to confirm during a pilot
Accounts and permissions
Build permission boundaries around role, team, model usage and data access, rather than having all members share one entry point.
Traceable process
Keep necessary records of the research process, backtest experiments, risk review and key conclusions to support team review and accountability.
Sensitive-action confirmation
For trade execution, strategy go-live, fund-related actions or external output, keep a human confirmation step rather than auto-executing directly.
Exception handling
When model output is unstable, data is missing or an external system fails, the workflow should degrade to human review and a clear pause mechanism.
Deployment
Deployment and pilot options
Public trial
Fits individuals trying the desktop, built-in agents, cloud models and the credit system; start with the public download and quick-start flow.
Team pilot
Fits connecting FiClaw into research, quant, risk-control or pre-trade check workflows; confirm the use case, people, data and budget boundaries first.
Enterprise integration
For internal systems, private data, compliance requirements or more complex deployment, evaluate interfaces, permissions, audit and operations responsibility separately.
Limits
Public-disclosure boundaries
- FiClaw does not promise investment returns and does not replace the professional judgment, compliance review or risk-control responsibility of a licensed institution.
- Model output may contain errors, omissions or outdated information; key conclusions in financial settings must go through human review.
- Public pages do not show unfinished, unsigned or non-disclosable client cases, system details or security configuration.
- Security, compliance and deployment plans keep evolving with product stage, client scenario and regulatory requirements.
Need to scope a team pilot?
If you plan to connect FiClaw into real research, quant, risk-control or pre-trade check workflows, map out data, permissions, review nodes and deployment constraints first.