SkillAgentSearch skills...

InvestorClaw

Portfolio analysis & market intelligence for Claude Code & OpenClaw. Deterministic Python math + Claude/OpenClaw synthesis. Educational guardrails, no investment advice.

Install / Use

claude mcp add argonautsystems -- npx -y github:argonautsystems/InvestorClaw

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

80/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of InvestorClaw

InvestorClaw scores 80/100 on our quality scale, 96th of 110 Finance & Accounting skills we index.

Its MCP Server is 19 KB long, well organised into 25 sections with 14 code examples: a thorough specification that gives an agent plenty to work with.

It has 10 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
4/20
Freshness
11/15

Maintenance, license and trust

  • The repository was last updated about 5 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
  • It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 95/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

InvestorClaw compared with similar skills

All 4 of these similar skills score higher than InvestorClaw; compare them before choosing.

SkillScoreStarsUpdatedFormat
InvestorClaw (this skill)by argonautsystems80105mo agoMCP Server
Agent-Reachby Panniantong100109.0ktodayCLAUDE.md
headroomby headroomlabs-ai10078.9ktodayCLAUDE.md
CowAgentby zhayujie10047.8ktodayCLAUDE.md
Scraplingby D4Vinci10097.5ktodayMCP Server

Frequently asked questions

How do I install InvestorClaw?
Run claude mcp add argonautsystems -- npx -y github:argonautsystems/InvestorClaw. The install tabs above show the steps for each supported agent.
Which AI agents does InvestorClaw work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is InvestorClaw safe to use?
It is Apache-2.0-licensed and scores 95/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is InvestorClaw still maintained?
The repository was last updated about 5 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.

InvestorClaw

skills.sh

<p align="center"> <picture> <source srcset="assets/investorclaw-logo.webp" type="image/webp"> <img src="assets/investorclaw-logo.jpg" alt="InvestorClaw" width="200"> </picture> </p> <p align="center"> Portfolio analysis and market intelligence | v4.5.0 | Apache 2.0 + MIT-0 | Educational Use Only </p>

InvestorClaw is a self-contained containerized software package that any MCP-capable agent calls into. It is not a markdown skill. It is not a prompt injection. It is the adapter and distribution layer for a real portfolio engine, packaged for agent runtimes that know how to speak MCP.

The deterministic engine lives in ic-engine. Foundation primitives live in clio. The runtime container lives in mnemos-ic-runtime.

New to InvestorClaw? The fastest path is the prebuilt container — see Getting Started: docker pull ghcr.io/argonautsystems/ic-engine:4.7.7-cpu, mount a portfolio CSV, done. No source build. Works with any market-data provider (free yfinance by default).

Optional Stonkmode adds live commentary from 30 fictional cable TV finance personalities. It is entertainment and education on top of the portfolio surface, not a requirement.


Features

InvestorClaw separates agent-facing commands from portfolio computation.

  • Containerized engine. The current runtime is a single Docker container with MCP-HTTP on port 18090 and the dashboard portal on port 18092. Start the container, point an MCP-capable agent at it, and ask portfolio questions.
  • Deterministic computation. There is no LLM in the parse path. Holdings parsing, normalization, analytics, and envelopes are deterministic. Narration is optional, provider-swappable, and sits on top of structured output.
  • Nine asset classes. InvestorClaw handles equities, ETFs, bonds, mutual funds, options, cash, CME futures (priced via Massive's /futures/vX feed with correct contract-multiplier notional), crypto, and precious metals.
  • Massive market data. Massive (massive.com) is the primary paid provider — batch quotes for large portfolios and CME futures via /futures/vX with correct contract-multiplier notional. Set MASSIVE_API_KEY and optionally INVESTORCLAW_PRICE_PROVIDER=massive. Full guide: Getting Started. Full provider matrix: docs/PROVIDERS.md.
  • Twenty MCP tools. The primary surface is portfolio_ask, portfolio_initialize, portfolio_initialize_status, portfolio_holdings, portfolio_refresh, portfolio_setup, portfolio_keys_status, portfolio_keys_set, portfolio_keys_delete, portfolio_keys_recommend, portfolio_keys_backup, portfolio_keys_restore, portfolio_keys_backups_list, portfolio_response_get, portfolio_response_list, portfolio_response_delete, portfolio_response_flag_bad, portfolio_version_check, portfolio_export, and portfolio_import.
  • 17-tab dashboard portal. Open http://localhost:18092 for Overview, Holdings, Performance, WhatChanged, Scenarios, Bonds, Optimize, Cashflow, Peer, Analyst, News, Markets, Lookup, Synthesis, Reports, Settings, and About.
  • Regenerate from the browser. The Overview tab has a Regenerate button that fires setup, refresh, and all 12 analyzers as a background sweep.
  • Web upload form. Settings accepts multipart uploads for .csv, .tsv, .xls, .xlsx, .pdf, .json, .ofx, and .qfx portfolio files.
  • Pluggable, battery-tuned narration. Narration is two-stage — a consultant compresses the signed envelope into a fact-faithful summary, then a narrator enriches it. The 30-prompt COBOL hallucination battery (harness/cobol/PROVIDER_HALLUCINATION_REPORT.md) settled the defaults: consultant = deepseek-v4-flash via the direct DeepSeek API (cheapest, and matches gemma-4-31B on hallucination with better coverage); narrator = gemini-3.1-pro (lowest fabrication) or llama-3.3-70b on Groq/Together (max coverage). Any OpenAI-compatible endpoint works — set INVESTORCLAW_CONSULTANT_* and INVESTORCLAW_NARRATIVE_* (see Getting Started).
  • Safe fallback defaults. InvestorClaw can start with no API keys and use the yfinance fallback. Optional keys improve news, ratings, and premium data coverage.
  • HMAC-signed envelopes. Outputs are tamper-evident. They are not encrypted.
  • No fabrication path. The engine returns what it can prove from data, marks gaps, and avoids pretending that missing facts exist.
  • No brokerage credentials. InvestorClaw does not ask for brokerage logins and does not need them.
  • No outbound trades. This package analyzes portfolios. It never places orders.
  • Cobol NLQ barrage coverage. The 30-prompt natural-language query barrage in harness/cobol/nlq-prompts.json currently stands at 29/30 PASS.

Native cross-runtime coverage varies.

See docs/AGENT-COMPARISON.md (coming soon) for the per-runtime provider matrix.

Note: Effective April 4, 2026, Anthropic Claude subscriptions (Pro at $20/mo and Max at $100–$200/mo) no longer cover use from third-party agent runtimes like OpenClaw, ZeroClaw, or Hermes Agent. The plan limits do not apply there, and a subscription alone will not authenticate those calls.

To use Claude models from a third-party agent, either enable pay-as-you-go "extra usage" billing on the subscription or connect with a direct API key on metered billing.

Claude Code is different. Anthropic subscriptions continue to apply normally there at standard rates.

Battery-validated narration stack (harness/cobol/PROVIDER_HALLUCINATION_REPORT.md): consultant deepseek-v4-flash (direct DeepSeek API — cheapest, matches Gemma-4-31B), narrator gemini-3.1-pro (lowest fabrication) or llama-3.3-70b via Groq/Together (coverage). Use Claude Code if you want Claude.

Sources: PYMNTS 2026-04-04, VentureBeat


Non-Goals

InvestorClaw helps you have informed conversations with your financial advisor by surfacing data-driven insights. It does not manage money. It does not execute trades. It does not provide investment advice.


Comparison With thinkorswim

InvestorClaw helps you understand your portfolio. thinkorswim helps you execute trades.

| | InvestorClaw | thinkorswim | |---|---|---| | Purpose | Portfolio analysis & insights | Active trading & execution | | Can trade? | ❌ No | ✅ Yes | | Data source | Free (yfinance) + optional paid (Massive, Finnhub) | Real-time (proprietary) | | Run locally? | ✅ Yes (Docker / yfinance fallback) | ❌ Cloud only | | Open source? | ✅ Yes (Apache 2.0) | ❌ Proprietary | | Target user | Individual investors + advisors | Professional/active traders | | Best for | Understanding your portfolio | Executing trades + charting |

Use InvestorClaw to analyze.

Use thinkorswim to execute.

Use both if you want analysis and execution in one workflow.


Quick Start

InvestorClaw supports Claws-family, Claude Code, Hermes Agent, and standalone Docker Compose deployment paths.

Choose the platform that matches how you work.

Claude Code / Claude Desktop

Add InvestorClaw as a plugin marketplace directly from this repo, then install:

/plugin marketplace add argonautsystems/InvestorClaw
/plugin install investorclaw

First question:

What are my current holdings?

No source clone, no uv, no Python installers. The plugin pulls the public ic-engine container and drives it over MCP-HTTP — the same skill and same container as every other runtime below. Full walkthrough, including how to pick a market-data provider: docs/GETTING_STARTED.md.


OpenClaw

OpenClaw fits Linux and macOS workstations and servers.

Install InvestorClaw as an OpenClaw skill:

clawhub install investorclaw

First question:

Run a full portfolio analysis.

ZeroClaw

ZeroClaw fits Raspberry Pi and other ARM devices.

Install InvestorClaw:

clawhub install investorclaw

First question:

What's my bond allocation?

Hermes Agent

Hermes Agent installs InvestorClaw inside the NousResearch agentic CLI.

Hermes Agent is the agentic CLI.

It is not a model.

You can pair it with any provider the agent supports. That includes cloud providers such as OpenAI, Together, xAI, OpenRouter, and Nous Portal. It also includes fully local providers such as Ollama, vLLM, llama-server, and LMStudio.

Note: The Hermes LLM family (Hermes 3, Hermes 4 - NousResearch fine-tunes of Llama/Qwen) is a separate product. Hermes Agent can use a Hermes LLM as its backend model, or any other model.

clawhub install investorclaw

First question:

Summarize my portfolio performance.

Standalone Docker Compose

Use the runtime container directly when you want InvestorClaw without an agent-specific installer.

Runtime source: ncz-os/mnemos-ic-runtime.

docker compose up -d

Dashboard: http://localhost:18092.

Then connect any MCP client to:

http://localhost:18090/mcp

First question to agent:

List available portfolio tools.

Dashboard / Web Portal

The local dashboard runs at http://localhost:18092.

It is the browser surface for people who want to inspect the portfolio, upload files, regenerate analytics, manage provider keys, and read saved reports without turning every action into a chat prompt.

| Tab | What it shows | |---|---| | Overview | Live portfolio snapshot plus the Regenerate button, which triggers the full 12-analyzer sweep | | Holdings | Position-level detail with cost basis, market value, and gain/loss | | Performance | Time-series return charts, alpha, beta, and Sharpe | | WhatChanged | Delta view between the last two refreshes | | Scenarios | What-if analysis for allocation shifts | | Bonds | Duration, yield, and credit quality breakdown | | Optimize | Mean-variance and risk-parity weight suggestions | | Cashflow | Dividend and coupon income calendar | | Peer | Relative performance vs benchmark and peer basket | | Analyst | Wall Street consensus estimates and ratings | | News | Recent headlines filtered to portfolio constituents | | Markets | Broad market context across indices, sectors, and macro | | Lookup | Symbol search and quick quote | | Synthesis | Narrative summary generated by the configured LLM provider | | Reports | Saved HTML reports and EOD email archives | | Settings | API key management, portfolio file upload, and provider config | | About | Version, license, and build info |

The Regenerate button on Overview fires setup, refresh, and all 12 analyzers as a background sweep. Results normally become visible in Overview within about 60 seconds.

The upload form on Settings sends a multipart POST and accepts .csv, .tsv, .xls, .xlsx, .pdf, .json, .ofx, and .qfx files.


Commands / MCP Tools

See docs/MCP_TOOLS_REFERENCE.md for the full command contract.

| Tool | Purpose | |---|---| | portfolio_ask | Natural-language query against current portfolio data | | portfolio_initialize | Seed the engine with uploaded holdings | | portfolio_initialize_status | Poll initialization progress | | portfolio_holdings | Return current ho

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars10
CategoryFinance
Updated4mo ago
Forks2

Languages

Python

Trust signals

95/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

2 info