qanat
OfficialUses DuckDB as the default local store for factor data and backtest results, allowing the MCP server to persist and query project data in a DuckDB file.
Provides a SQL connector for reading from any database reachable through SQLAlchemy, enabling ingestion from SQL data sources.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Why FidetoLabs built this
You are thinking about buying a company. You go and read about it. There are posts explaining why it is going up, a video from someone who sounds like they know, charts, and a summary of last quarter that reads like a professional wrote it.
Any of that now takes a minute to make and costs nothing. Some of it is wrong, and you cannot tell which.
The filing is still on EDGAR. It still says what the company told its regulator, and reading it takes as long as it always did.
FidetoLabs cannot make lying expensive again. It can make checking cheap. Qanat is the first tool for that: whatever number it gives you, it can show you the filing the number came out of and the day that filing was published.
You write the logic and Qanat runs the calculator. The formula, the threshold and the list of companies are yours. It tells you what came out and never tells you what to do about it.
Related MCP server: lab-mcp
What Qanat is
Qanat reads company filings, computes the metrics you define, and answers as of any past date.
A company publishes its results weeks or months after the period they cover, so every number here carries two dates: that period, and the day it became public. Ask about March 2022 and you get what was public in March 2022, not what we know now.
Your agent is the interface.
Not investment advice. What you do with the number is your decision and your risk.
Getting started
uv tool install qanat-fdtl # or: pip install qanat-fdtl
export SEC_USER_AGENT="your name you@example.com"
qanat init my-research --demo && cd my-research
claude mcp add qanat -- qanat mcp --scope researchThe demo pulls a few real 10-K filings from EDGAR and reads the numbers out of
them, in under a minute. No API key and no account: the SEC only wants a
contact address. Any MCP client works, with qanat mcp as the command and the
project folder as its working directory.
Using it
Ask your agent. It can find filings on EDGAR, pull the ones you choose, define a metric as a formula, and screen on it as of a date or across a window.
"What data is in here, and where did it come from?"
"Pull Apple and Microsoft's last five annual reports."
"Define net margin, then show me who passed it in 2022 and what they did next."
--scope decides how much it can do. Four levels, each including the one
before: data reads what is already there, research adds finding filings and
screening, full adds ingesting and running, authoring adds defining metrics
and changing the pipeline. Pick the smallest that does your job.
full is the default, and authoring sits above it so an agent screening
filings cannot rewire the pipeline by accident.
examples/sec-10k
is a corpus of real filings to try it on.
docs/dataflow.md
walks the pipeline end to end,
docs/agents.md
is the tool surface, and
docs/flows.md,
docs/contract.md
and
docs/words.md
cover why each tool exists, the pipeline rules and the vocabulary.
The limits
It is beta. It works for my own use and few other people have run it. If something breaks, open an issue or say so on Discord.
It cannot place an order, and no agent it serves ever will.
Nothing is added up. No average across years, no equity curve. That needs a decision about how much of each company to hold, and the decision is yours.
The dates protect a transform, not you. It cannot see past its date; you and your agent can, while you are deciding what to try.
Two sources: SEC EDGAR for filings and Yahoo Finance for prices. Only the 10-K and the 10-Q are parsed; other forms are found and labelled, not read.
For general pipelines, use something else. Airflow, Dagster and Prefect run any DAG across many machines. Qanat runs one shape.
Pull requests welcome. See CONTRIBUTING.md. MIT licensed. Copyright (c) 2026 fidetolabs.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Build, test and publish agentic workflows from your AI client: approve steps, read tables, see runs.
Workflow planning, recovery checkpoints, coordination, fixtures, and compatibility tools for agents.
Build, backtest, and deploy quantitative trading strategies from your AI agent.
Design, build and run teams of AI agents with checked handoffs between them.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables users to create and revise strategy graphs, modules, and custom nodes; run exact-revision backtests; and maintain project knowledge and artifact documentation.MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to run a full quant research workflow over MCP: pulling data, authoring and backtesting strategies, running statistical validation and risk checks, and recording findings for future sessions.1MIT
- AlicenseNot gradedqualityBmaintenanceEnables agents to search and compose typed operators into validated, runnable pipelines, propose missing operators or types as drafts, and crystallize proven drafts into stable operators through a human-reviewed GitHub pull request.Apache 2.0
- AlicenseNot gradedqualityCmaintenanceEnables users to prepare and validate market data, launch and control backtests, and retrieve results, causal explanations, and self-contained reports through stdio tools. It also supports polling or canceling jobs and running sweep, walk-forward, factor, and study research workflows with bounded pagination and reproducible artifacts.MIT