ClearFox logoClearFox
Architecture & data flow

How ClearFox works

From question to insight in seconds. Here's exactly how ClearFox connects your tools and delivers answers to your team.

System architecture

Your team

CEO, CFO, COO, VP's, Team Leads

Ask questions in natural language

Your infrastructure
ClearFox portal

Web interface, auth, RBAC, chat history, usage limits

AI engine

Orchestrates tool calls, builds context, streams answers

MCP connectors

30+ isolated Docker containers, one per tool

Secure API calls (read & controlled actions)
CRM & Sales
Salesforce, HubSpot
Finance
Stripe, QuickBooks
Project Mgmt
Jira, GitLab
Databases
MySQL, PostgreSQL
Analytics
GA4, Power BI
Support
Zendesk, Freshservice
HR
BambooHR, Personio
Communication
Slack, Mailchimp
Cloud models: Anthropic Claude or OpenAI. Your query context is sent via API — never raw data. You bring your own key.
Local models: Ollama or vLLM on your hardware. Nothing leaves your network. Full air-gap support.

What happens when you ask a question

The entire process takes 10–30 seconds and is fully transparent. You can watch every step in real time.

01

You ask a question in plain language

"What are our biggest delivery bottlenecks and what financial risks should I know about this quarter?"

02

The AI plans its approach

It understands your question, checks which data sources are available for your role, and plans a sequence of tool calls.

03

Connectors query your tools

Each connector queries its tool via API: Jira for tickets, Stripe for payments, GitLab for code metrics, Salesforce for pipeline — in parallel where possible.

04

Results are synthesized

The AI gets structured data from each tool, cross-references it, finds patterns, and builds an answer with specific numbers and actionable recommendations.

05

You get a clear answer

The analysis shows up in your chat — with references to specific data sources. Ask follow-up questions to drill deeper into any finding.

Powered by the Model Context Protocol (MCP)

MCP is an open standard that enables AI models to interact with external tools safely and predictably. Works with any model — Anthropic, OpenAI, or local.

One connector per tool

Each business tool has a dedicated Docker container running its MCP connector. Isolated, secure, independently updatable.

Read and act

Connectors can read data and take actions — create tickets, send messages, open merge requests. You control which actions are allowed per role.

Up to 100 tool calls per query

The AI chains dozens of calls across multiple tools to build a deep, multi-source analysis — automatically.

Deploy in three steps

No ML expertise needed. No cloud dependencies. Just Docker and 5 minutes.

01

Deploy with Docker

One command spins up the portal, database, and all connectors.

docker compose up -d
02

Connect your tools

Add API keys in the admin panel. Set up SSO, create roles, and assign data sources to each team. ClearFox handles the rest.

03

Start asking questions

Invite your team. Everyone gets role-appropriate access. Ask anything — from quick metrics to deep cross-system analysis.

Stop guessing. Start knowing.

Give your leadership team an AI-powered analyst that detects bottlenecks, identifies risks, and delivers data-driven insights — in seconds.