Jarvis is a sophisticated AI assistant designed specifically for Independent Financial Advisors (IFAs). Built on a "Board of Specialists" architecture, Jarvis autonomously monitors client data, identifies risks, and drafts professional communications.
This final submission significantly elevates the preliminary architecture to directly address the Final Round Evaluation Criteria:
- "MCP Integration Backbone" (Technical Ambition): Replaced static Python tools with true Model Context Protocol (MCP) servers. Jarvis now natively integrates with a Calendar MCP (for scheduling and email management) and a Market Feed MCP (for live macroeconomic data and news).
- Action Execution - "Adviser approves, not produces" (30% Adviser Would Use It): Shifted the UI paradigm from passive notifications to active staging. When Jarvis's heartbeat detects an issue, the
EmmaandColinsub-agents proactively draft a compliance-checked response. The advisor simply clicks "Approve & Send" or "Edit" in the dashboard, ensuring complete regulatory control. - Pre-Meeting Insights: Integrated a "Get Insights with Atlas" feature directly into the Calendar UI. Before a meeting, Atlas automatically queries ChromaDB to map the client's past discussions, risk tolerance, and open action items into a pre-meeting brief.
- Multi-Channel & Voice (25% Technical Ambition): Pushed the boundaries of the tech stack by implementing Telegram integration with agent-routing (
@jarvis,@emma,@colin) and a WebSocket-based Voice Pipeline (Sandwich Architecture) featuring real-time STT and TTS.
View Demo Videos and Slide Deck on Google Drive
| Component | URL |
|---|---|
| Frontend | advisory-ai-jarvis.vercel.app |
| Backend API | Railway API Docs |
Login Credentials: Username:
abimanyu| Password:admin
- Jarvis Orchestrator: The central brain that manages delegation and identity synthesis.
- Atlas (RAG Specialist): Deep-dives into client files, emails, and transcripts.
- Emma (Paraplanner): Drafts suitability reports and advisor correspondence.
- Colin (Compliance): Ensures all outputs meet UK FCA regulations.
| Context | Model |
|---|---|
| Direct Chat (Jarvis) | openai:gpt-5.1-chat-latest |
| Heartbeat & Cron Jobs | openai:gpt-4o-mini |
| Voice Agents (TTS & STT) | Built-in OpenAI models |
Jarvis is built using LangChain Deep Agents, a framework for creating sophisticated AI agents with tool use, planning, and sub-agent delegation. This provides Jarvis with:
- Autonomous reasoning with multi-step planning.
- Tool orchestration with built-in filesystem, web, and custom tools.
- Sub-agent delegation for specialized tasks.
- Telegram Integration: Advisors can communicate with Jarvis or any of the sub-agents directly from Telegram, making the assistant accessible on the go.
- Voice Agents: Jarvis and its sub-agents feature built-in voice capabilities. Text-to-Speech (TTS) and Speech-to-Text (STT) are seamlessly integrated using OpenAI models, enabling natural Voice-to-Voice interactions via a "Sandwich Architecture".
Instead of a static prompt, Jarvis builds its personality and context dynamically at runtime by reading workspace files:
SOUL.md- Core personality and values.IDENTITY.md- Professional identity as a financial advisor assistant.USER.md- Information about the advisor Jarvis serves.HEARTBEAT.md- Instructions for autonomous background checks.
Jarvis delegates specialized tasks to three expert sub-agents:
| Agent | Role | When Used |
|---|---|---|
| Atlas | RAG Specialist | Searches client documents, emails, and transcripts |
| Emma | Paraplanner | Drafts professional client communications |
| Colin | Compliance | Validates outputs against UK FCA regulations |
Jarvis natively connects to standardized Model Context Protocol (MCP) servers:
- Calendar MCP: Directly integrates calendar capabilities. The UI features a dedicated Calendar Page, where advisors can click "Get Insights with Atlas" on any meeting to automatically have Atlas query relevant client files and prepare insights.
- Market News MCP: Equips the orchestrator with specialized tools to fetch macroeconomic indicators, financial news, and asset performance via Tavily and other sources.
Jarvis and Emma autonomously draft responses to incoming client emails.
- These drafts are populated directly in the UI's Email Drafts section.
- The advisor can review, edit, approve, or reject these drafts before they are sent, ensuring complete compliance and control over client correspondence.
Jarvis runs continuously in the background using APScheduler.
- During a heartbeat, Jarvis monitors live market news and client files.
- Direct Actions: If urgent info is found, Jarvis can directly use its tools to Raise a Notification to the advisor via the UI/Telegram, or generate an Email Draft to proactively handle the situation without explicit prompting.
Client documents (emails, transcripts, policies) are embedded and stored in ChromaDB for semantic search. This enables:
- Context-aware retrieval - Find relevant documents based on meaning, not just keywords
- Cross-client analysis - Search across all clients for similar situations
Jarvis has access to both default and custom tools:
Default Deep Agent Tools:
- Filesystem operations (read, write, list, glob)
- Web browsing and search
- Planning and task management
Custom Financial & System Tools:
| Tool | Purpose |
|---|---|
get_macro_indicators |
Fetches macroeconomic indicators via Market News MCP |
search_financial_news |
Searches for recent UK Financial/Regulatory news via Market News MCP |
get_asset_performance |
Checks performance snapshots for assets via Market News MCP |
raise_notification |
Sends an urgent alert directly to the dashboard from the heartbeat |
draft_email |
Drafts a client email for advisor review |
retrieve_context |
Vector search in ChromaDB (Atlas) |
search_uk_compliance |
FCA regulatory search (Colin) |
- Username:
abimanyu - Password:
admin
Test Jarvis's ability to scan his entire "book" for urgent matters.
- Launch the Dashboard: Run
cd frontend && npm run dev. - Open the Chat: Navigate to the "Chat" page.
- Ask the Query: "Show me anything in the last 10 days that looks urgent across my book (emails and meeting notes)?"
- Behind the Scenes: Jarvis will scan
datasets/**for files modified between Jan 28 and Feb 08, 2026, and use Atlas to identify risks.
Test how Jarvis identifies new incoming data and alerts you autonomously.
- Prepare the Data: Locate the sample email in
sample/2026-02-01_ill_situation.txt. - Upload: Go to the "Clients" page in the dashboard -> Select Gareth Cheeseman -> Upload the sample file as an "Email Archive" document.
- The Result: Within 30 minutes (or on the next heartbeat), a notification or email draft will appear in the Dashboard: "🚨 Jarvis Alert: Gareth Cheeseman has emailed regarding income protection policies due to illness..."
Test how advisors can talk to any of the sub-agents on the go.
- Setup Token: Ensure
TELEGRAM_BOT_TOKENis set in your.envfile. - Start the Bot: Open a terminal and run
uv run python src/jarvis/telegram_bot.py. - Chat on Telegram: Go to telegram and find your bot.
- Invoke Agents: Prefix your messages with the agent's name to route the query:
@jarvis Who are my top 3 clients by AUM?@emma Draft a quick email to Gareth letting him know we are looking into his policy.@atlas What did Alan say in our last meeting about his risk tolerance?
python scripts/ingest_documents.py
uv run uvicorn jarvis.api:app --reload --port 8000
Note: The heartbeat scheduler now runs as a background thread within the API. Websockets are enabled and expose the STT/TTS sandwich layers.
cd frontend && npm run dev
src/jarvis/: Core Agent and Sub-agent logic.workspace/: Operating environment, prompts, and client datasets.frontend/: Advisor dashboard (React + Vite).sample/: Standardized text data for system validation.skills/: Modular skill definitions.
The ultimate execution of the "Proactive Hunch Engine." Jarvis will serve as a background listener during Zoom/GMeet client calls. By processing streaming audio transcripts in real-time, Jarvis will push hidden, compliance-checked prompts to the advisor's dashboard mid-conversation (e.g., "Client mentioned retiring early. Their current trajectory won't meet the £55k target. Suggest modeling a salary sacrifice.").
Embed Jarvis in WhatsApp or Slack group chats between the advisor and client families. To maintain strict FCA compliance, Colin will audit every response generated by Jarvis. The advisor will use an "Approve & Forward" workflow, ensuring clients get rapid, accurate answers while keeping the advisor fully in the loop.
Evaluate Jarvis quality using human judges (advisor ratings on response quality), LLM-as-judge scoring for consistency, and task completion metrics.
"I'm not just a chatbot; I'm a teammate who never sleeps." — Jarvis