Cloud model routing
FiClawCloud centrally manages the model catalog, model selection, call consumption and credit records, so the desktop app avoids wiring up multiple model providers directly.
Integrations
FiClaw integrations
When a financial agent enters business processes, beyond capability itself, you need clarity on how models, data, interfaces, permissions and audit connect to the workflow safely.
Scope
What the access layer covers
Quant capability interface
FiClawQuantAPI unifies quote, financials, backtest, paper-trading and risk-control capabilities, so agents call capabilities inside a workflow instead of touching underlying systems directly.
Institutional data sources
During a team pilot, you can assess the access boundary for public material, internal material, research documents, rule bases and structured data based on the scenario.
Collaboration & lead systems
The site's Contact / Demo forms currently write to Feishu Bitable via a server endpoint and notify an internal group; later team integration is assessed separately per client process.
Permissions & audit
Where internal systems or sensitive data are involved, account roles, data permissions, action logs, human confirmation and exception handling must be defined explicitly.
Workflow skills
FiClawSkills wraps data, model and tool capabilities into reusable tasks, so a financial agent can enter research, quant and risk-control flows by role.
Modes
Three access modes
01
Public trial
Use FiClaw Desktop, cloud models and public capabilities without connecting internal systems — good for first experiencing roles and model routing.
02
Light pilot
Connect limited data, rules or documents around a single business scenario, focusing on task flow, review traceability and output quality.
03
Custom integration
When internal systems, private data, permission audit or deployment constraints are involved, assess interfaces, data boundaries and operations responsibility separately.
Principles
Access principles
Define the business scenario first, then decide which data and tools to connect.
Connect only necessary data first; do not expose unrelated systems to the agent at once.
Keep human-review nodes first, then gradually assess the automation boundary.
Record key process first, then judge whether to extend to more teams and systems.
Boundary
Public boundaries
- The site does not publish internal interface docs, keys, backend paths or client system details.
- ficlaw-data is an internal read-side data service, not a direct access entry for site users.
- Actions involving live trading, funds, client assets or external disclosure must keep human confirmation and compliance boundaries.
- Each team's data access scope, retention, model context and audit requirements must be confirmed separately before a pilot.
Need to assess data or interface access?
If your team plans to connect internal data, research documents, rule bases, quant interfaces or existing systems, start with a boundary-clear team pilot.