Agentic trading · MCP research layer
Agentic trading needs a research layer
Your broker's AI agent can place trades. It can't tell you what's worth buying. ThesisLoop's MCP server gives any AI trading agent the research layer — screen 5,000+ stocks, build cited theses, interrogate filings, and monitor positions — grounded in company filings. ThesisLoop never places trades.
The shift
What is agentic trading?
Agentic trading is when an AI agent — not a person clicking buttons — proposes and places trades on your behalf. Through 2026, brokers began opening MCP endpoints so an agent can connect to a trading account with a single line in its configuration.
Robinhood's Agentic Trading is the flagship example: AI agents connect over MCP to a dedicated Agentic Account with a budgeted wallet, per-trade notifications, and a one-tap disconnect. The beta supports stocks, with options, crypto, and futures planned.
The difference
Agentic trading vs. algorithmic trading
Algorithmic trading is rule-based execution. A human encodes a strategy up front — price triggers, timing, volume, factor signals — backtests it, and the system fires orders whenever the conditions match. The rules never change mid-flight, and the system never reads a filing: if the thesis behind the rules breaks, the algorithm keeps trading it anyway.
An AI trading agent works from a goal, not a rule set. Given a mandate — “research my watchlist and propose what's worth holding” — it decides which tools to call over MCP: screen a catalog, build a thesis, interrogate the latest filings, check whether management delivered on its promises. It reasons over the evidence, then proposes an action, and it can revise its view when new information lands. Copy trading, for contrast, skips the reasoning entirely — it mirrors someone else's positions without doing any research of its own.
A working agentic trading stack therefore has three parts: an MCP-capable AI agent that reasons and decides, a research layer that grounds those decisions in cited primary sources, and an execution layer at your broker with budgets and controls. ThesisLoop is the research layer in that stack — it informs the decision and never touches execution.
The gap
Execution without research is just noise
An execution MCP is very good at one thing: placing the order it is told to place. It does not tell the agent what is worth buying, whether management has delivered on its promises, or which risks are hiding in the latest filing.
Execution platforms are explicit about this. Their own disclosures warn that AI agents “can make errors, misinterpret instructions, act on incomplete or outdated information,” and that the broker “does not control, supervise, monitor, recommend, or audit these AI agents.” That is exactly the gap a research layer fills: give the agent grounded, cited research before it decides.
The architecture
How ThesisLoop fits
Three nodes, one clean separation of duties. Your agent asks ThesisLoop what's worth doing, then asks your broker to do it — within the budget you set. ThesisLoop is research-only and never touches your brokerage balance.
Your AI agent
Claude, Cursor, Codex, or any MCP-capable client — the brain that decides.
ThesisLoop MCP · research
Screen, build theses, run analysis, interrogate filings, monitor positions. Never places trades.
Your broker's MCP · execution
Places and settles orders within your budgeted account and controls.
The tool surface
What your AI trading agent can do
ThesisLoop's MCP exposes its full research surface as agent tools, grouped here in plain English.
Screen & discover
Point the agent at the market and let it find candidates worth a closer look.
- Search a catalog of 5,000+ stocks across the US and India
Build & analyze theses
Turn a shortlist into a structured, cited investment thesis with full AI analysis.
- List, read, and export existing theses
- Create a new thesis from selected stocks
- Run full AI analysis and check its status
Interrogate filings
Ask questions of the primary source and get answers cited to the page.
- Check and ensure a company is indexed for Q&A
- Get suggested questions worth asking
- Ask a question and get a cited answer
- Review question history
Monitor positions
Keep the thesis honest after the trade — flag when the story changes.
- Set up, list, create, or update monitoring loops
- Pause, resume, or trigger a loop now
- Read run history and material updates
Setup
Set up in two minutes
Create an account, mint a token, paste one config block, and your agent can research before it trades.
Create a free account
Sign up for ThesisLoop. MCP access comes with the account — analysis quotas follow your plan.
Mint your MCP token
Open /dashboard/mcp-access and generate a token (prefixed tl_mcp_). Pick the read-only preset or the research-and-loops preset.
Paste the MCP config
Add one server entry to your agent's MCP config with the ThesisLoop URL and your Bearer token.
Research before you trade
Tell your agent to pull the ThesisLoop thesis, conviction, and filing risks before it proposes a single order.
agent MCP config
tokens minted at /dashboard/mcp-access
{
"mcpServers": {
"thesisloop": {
"url": "https://thesisloop.ai/api/mcp",
"headers": {
"Authorization": "Bearer tl_mcp_••••••••"
}
}
}
}In practice
Example agent prompts
Wire ThesisLoop into your agent's workflow so research always comes before the order.
“Before you buy anything today, pull the ThesisLoop thesis and conviction score for each ticker on my watchlist and summarize the top three risks from the latest concall.”
“Screen the ThesisLoop catalog for semiconductor names, create a thesis on the two with the strongest business-model scores, and run full analysis.”
“Ask ThesisLoop's filings Q&A whether management delivered on last year's capex guidance, with citations, before you size any position.”
“Set a weekly ThesisLoop monitoring loop on my holdings and alert me the moment a filing or news item changes the thesis.”
FAQ
Agentic trading questions
What is agentic trading?
Agentic trading is when an AI agent — not a person clicking buttons — proposes and places trades on your behalf. In 2026, brokers began opening MCP (Model Context Protocol) endpoints so agents can connect to a trading account with a single configuration line. Robinhood's Agentic Trading is the flagship example: an agent connects to a budgeted account and can place trades within limits you set.
Does ThesisLoop execute trades?
No. ThesisLoop is research-only. It never places, routes, or settles orders and never touches your brokerage balance. Its MCP server gives your agent the research layer — screening, theses, full AI analysis, filings Q&A, and monitoring — so the agent decides what is worth trading. Execution always happens on your broker's side, under your broker's controls.
How do I connect ThesisLoop to my AI agent?
Create a free ThesisLoop account, mint an MCP token at /dashboard/mcp-access, and add one server entry to your agent's MCP config with the URL https://thesisloop.ai/api/mcp and an Authorization header of 'Bearer' plus your token (tokens are prefixed tl_mcp_). Your agent can then call ThesisLoop's research tools. You choose a read-only preset or a research-and-loops preset when you mint the token.
Which AI trading agents and clients work with ThesisLoop?
Any MCP-capable client works — Claude, Cursor, Codex, and other agents that support remote MCP servers over Bearer-token auth. If your agent can add an MCP server by URL, it can use ThesisLoop as its research layer.
Can I use ThesisLoop with Robinhood Agentic Trading?
Yes — ThesisLoop supplies the research while your Robinhood Agentic Account handles execution. Follow the step-by-step guide at https://thesisloop.ai/agentic-trading/robinhood to connect both MCPs. Robinhood is a trademark of Robinhood Markets, Inc. ThesisLoop is not affiliated with, endorsed by, or sponsored by Robinhood.
What can my agent actually do through ThesisLoop's MCP?
The MCP exposes ThesisLoop's full research surface: search a catalog of 5,000+ stocks; list, read, create, and export theses; run full AI analysis and check its status; interrogate company filings with cited Q&A (index status, ensure indexed, suggested questions, ask a question, review history); and manage monitoring loops (set up, list, create or update, pause or resume, trigger now, and read run and update history).
What data does ThesisLoop's research use?
Research is grounded in primary company filings — annual reports, concall transcripts, and investor presentations for India, and SEC filings for the US. Findings are cited back to the source document so your agent works from evidence, not vibes.
Is agentic trading safe?
AI agents can make errors, misinterpret instructions, and act on incomplete or outdated information — the execution platform's own disclosures say as much. Keep a budgeted wallet, per-trade notifications, and a human in the loop, and use ThesisLoop's read-only token preset if you only want the agent to research and never modify anything. AI-generated research for informational and educational purposes only — not personalized investment advice, a stock recommendation, or a guarantee of returns. ThesisLoop is not a registered investment adviser. Do your own diligence and consult a licensed professional before making any investment decision.
How much does ThesisLoop's MCP cost?
MCP access comes with a ThesisLoop account. Analysis and Q&A quotas follow your plan, exactly as they do in the app — using the MCP does not unlock a separate or unlimited tier. You can start on the free plan and upgrade when you need more analyses.
AI-generated research for informational and educational purposes only — not personalized investment advice, a stock recommendation, or a guarantee of returns. ThesisLoop is not a registered investment adviser. Do your own diligence and consult a licensed professional before making any investment decision.
Robinhood is a trademark of Robinhood Markets, Inc. ThesisLoop is not affiliated with, endorsed by, or sponsored by Robinhood.