I build Canli Capital: open quant research where every published number comes with the file that produced it, including the strategies that failed.
Its flagship, canli-mcp, puts 306
free finance tools behind three for Claude and any MCP client: SEC filings, insider trades, 13F and
13D/G holders, options, FOMC and macro data, quant analytics and backtest validation. MIT, no API key,
runs on your machine: npx -y canli-mcp.
canlicapital.com · traceaxiom.com · ORCID 0009-0004-4138-7907
- Found that EDGAR's submissions API gives the same filing two different acceptance times, depending on which company's file you read: 450 filings measured, 13 of 25 files late by one Eastern offset (write-up, data on edgartools#1501).
- Fixed a look-ahead bug in lumibot's point-in-time SEC fundamentals: a 10-K accepted at 16:08 ET was visible to a backtest at 10:00 ET that day (#1209).
- canli-mcp 0.3.0: 306 finance tools for 1,109 tokens of context, measured against OpenBB, EdgarTools and Yahoo.
- Found that the Python package arch's SPA test, as merged in #871, rejects 13.5% of skill-less strategy searches at a nominal 5% (#879), and proposed a fix that brings it to about 6% (#881).
- The Null Zoo: how often backtest-overfitting corrections keep their promise, measured on synthetic searches where the truth is known.
- A placebo test for trading pipelines: 300,000 simulated tests; a t-test on the best rule said "edge" up to 78.9% of the time, the placebo held between 4.6% and 5.5% at a nominal 5%.
If any of this saves you from trusting a backtest that isn't real, a ⭐ on canlicapital helps other quants find it.
| Repository | What it does | Try it |
|---|---|---|
| canli-mcp | Every Canli server in one: 306 finance tools behind three (find, describe, run). SEC filings, ownership, options, the Fed and macro, quant analytics, backtest validation. | npx -y canli-mcp |
| canlicapital | The open research site, the free validation API and every MCP server, in one repository. | canlicapital.com |
| canli-validation-mcp | Is your backtest real? Deflated Sharpe, overfitting (CSCV), data-snooping tests and a backtest lab, with signed receipts. On the official MCP registry. | npx -y canli-validation-mcp |
| canli-research-mcp | Has this quant idea been tried, and how did it end? Papers, killed candidates and trial counts. | npx -y canli-research-mcp |
| canli-fundamentals-mcp | SEC fundamentals point in time: what a company first reported, and what was known on any date. | npx -y canli-fundamentals-mcp |
Want to help? The good first issues are small, real tasks, and every merged contribution is credited by name.
| Project | What it is |
|---|---|
| ALPHAC | The research engine behind every number on the site: validation framework, execution simulation, portfolio construction, and machine-readable evidence. |
| canli-backtest | A backtester where the obvious ways to fool yourself raise an exception: fill-time causality, execution realism, and multiple-testing accounting. |
| canli-pit-lake | A point-in-time market data lake that cannot accidentally tell you the future. |
| null-zoo | The Null Zoo: size and power of backtest-overfitting corrections (deflated Sharpe, haircut, bootstrap) on synthetic searches with known ground truth. |
| Validation API | The free HTTP API behind the MCP server: get a key, validate a backtest, keep a receipt. |
| validate-backtest-action | GitHub Action that validates a backtest through the API and adds a receipt badge. |
Each server is measured against the best alternative before release. The collection, factory and quality gate live in mcp-factory.
| Server | What it answers |
|---|---|
| actions-check-mcp | Checks GitHub Actions workflows: outdated actions, old Node runtimes, retired runners, injection. |
| citation-check-mcp | Are these citations real? Finds fabricated or mismatched references and retractions, returns clean BibTeX. |
| config-check-mcp | Validates config files against their official schemas: tsconfig, compose, workflows, 1,400+ more. |
| contact-check-mcp | Validates and formats phone numbers, email addresses and postal addresses for any country. |
| cron-check-mcp | Explains cron expressions, lists next run times in any time zone, converts between cron dialects. |
| dockerfile-check-mcp | Checks Dockerfiles: build-breaking mistakes, base image tags that exist, digests, platforms, EOL. |
| domain-health-mcp | Email and domain checks: SPF lookup limits, DKIM keys, DMARC, DNS records, registration expiry. |
| drug-label-mcp | FDA drug label answers with section citations, RxNorm name resolution, recalls and shortages. |
| end-of-life-mcp | Is this version still supported? End-of-life dates, latest patch and upgrade target for 470+ products. |
| internet-standards-mcp | RFC sections, status, obsoleted-by chains, errata and IANA registries. |
| kube-check-mcp | Checks Kubernetes manifests for your version: removed APIs, unknown fields, Pod Security, risks. |
| license-check-mcp | Open source license answers: SPDX ids, copyleft, and whether a dependency's license fits yours. |
| package-truth-mcp | Does this package exist? Version, deprecation, vulnerabilities and licence across 7 ecosystems. |
| recall-check-mcp | One recall check across CPSC, FDA and NHTSA, by name, model number, UPC or VIN. |
| regex-check-mcp | Tests regexes on real engines: JavaScript, Python, PCRE2, RE2 (Go). Matches, groups, ReDoS. |
| release-notes-mcp | What changed between two versions of a package: breaking changes, deprecations, security fixes. |
| satellite-imagery-mcp | The clearest Sentinel-2, Landsat, Sentinel-1 or NAIP scene for any place, with band links. |
| sql-check-mcp | Runs SQL on real PostgreSQL and SQLite in memory: your schema, the database's own errors, results. |
| vuln-priority-mcp | Which vulnerabilities to fix first: CISA KEV, EPSS, CVSS and SSVC in one ranking. |
| web-reader-mcp | Reads web pages and PDFs as clean Markdown: main content, the sections that answer a query. |
| world-time-mcp | Time anywhere, DST-safe conversions, holidays for 200+ countries, business days and meeting slots. |
| Project | What it is |
|---|---|
| TraceAxiom | A verification-first AI software engineering system that binds every accepted change to independent build, behaviour, accessibility, security and authority evidence. |
| Project | What it is |
|---|---|
| ArhanPassant (source) | A UCI chess engine and Rust chess library that keeps getting stronger: self-play, NNUE, and a calibrated SPRT gate for every new version. |
| Cubeduel (source) | Free, open-source speedcubing site: move-by-move solve review, server-verified ranked solves, smart cube support and a from-scratch Kociemba solver. |
| Sightduel | Ranked piano sight-reading. |
| Hollowtide | A survivor-like for HTML5 portals: deterministic simulation, no asset files, and a measurement harness. |
| Project | What it is |
|---|---|
| Longview | Research that cannot rewrite its record: a signed registry of calls. |
| PolyEdge | A second opinion on the real price of a Polymarket or Kalshi market. |
- Walk-forward and purged validation before admission claims
- Point-in-time lineage and untouched holdouts
- Deflated Sharpe, PBO, sensitivity, capacity, and regime tests
- Realistic costs, liquidity constraints, market status, and operational failure modes
- Trial ledgers, kill criteria, corrections, and negative-result publication
ALPHAC and Canli Capital publish research and paper-trading evidence. Nothing in these repositories is investment advice, an offer, or a representation of guaranteed performance.