SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI agentsOpenBusiness

OpenBusiness

OpenBusiness is an open-source Python command-line tool for business model research. It takes a company name, domain, and optional stock ticker, collects public evidence through channels such as Tavily search, Firecrawl scraping, and SEC EDGAR, then runs analysis via LangGraph-based analyst agents, ultimately generating a business model report in Markdown format. Unlike generic LLM questioning, OpenBusiness labels every claim in the report as "verified", "inferred", or "missing", and stress-tests assumptions that could change the analysis conclusions. It is aimed at founders, investors, consultants, product teams, and researchers, providing a quick structured preliminary report before in-depth due diligence. The tool supports OpenAI, Anthropic, and DeepSeek providers, supports report generation in English and Simplified Chinese, and outputs reproducible run artifacts.

Open website

About this tool

Purpose

OpenBusiness is an open-source Python command-line tool for business model research. It takes a company name, domain, and optional stock ticker, collects public evidence, runs LangGraph-based analyst agents, and generates a business model report in Markdown format. It is aimed at founders, investors, consultants, product teams, and researchers, and is suitable for quickly obtaining a structured preliminary company report before in-depth due diligence.

Its core product principle is: clearly state what is known, infer only when evidence supports it, and flag unknown information rather than hide it. Every claim is labeled as "verified", "inferred", or "missing", and this evidence-labeling contract runs through collection, analysis, synthesis, stress testing, and final drafting.

It is inspired by TradingAgents, but is not a bull-bear debate system; business model analysis requires decomposition, evidence tracing, and hypothesis stress testing, not a buy or sell vote.

Main Capabilities

  • A LangGraph-based multi-agent business analysis pipeline.
  • Evidence collection through Tavily search, Firecrawl scraping, and SEC EDGAR.
  • Outputs a business model canvas with explicit evidence labels: verified, inferred, or missing.
  • Built-in domain analysis packs: SaaS, public companies, e-commerce, AI infrastructure, consumer apps, investment screening, and general company research.
  • Built-in audience templates: standard analysis, investor memo, founder imitation research, competitor map, market research, and decision memo.
  • Reproducible run artifacts, including stage outputs, source extraction, and run metadata.
  • Through --language en and --language zh supports bilingual report generation.
  • Supports OpenAI, Anthropic, and DeepSeek providers.
  • A local configuration wizard, with hidden API key input and 0600 configuration file permissions.
  • Gives a language purity warning when generated reports mix output languages.
  • Pure Markdown reports, which can be directly archived, edited, or shared.

Usage

Install:pipx install openbusiness

One-off run:uvx openbusiness analyze "Notion" --domain notion.so

The installer will first let you choose English or Chinese, check Python, create a virtual environment if necessary, install the package in editable mode, and launch the configuration wizard in the selected language. The first language choice becomes the default terminal interface language as well as the default report language for openbusiness analyze , unless explicitly overridden later.

If you accept the default .venv setting, the installer will activate it during the installation session. Run it in each new terminal openbusiness needs to be activated again before. If you refuse to create a virtual environment, skip the activation step and use the Python environment where the package is installed.

Configure or reconfigure: the installer usually starts this wizard automatically. When the wizard is skipped, the terminal is switched before setup is complete, or the key needs to be updated later, it can be run manually. The configuration is saved in ~/.config/openbusiness/config.toml.

Analysis options:openbusiness analyze does not ask for language selection at startup. The terminal interface language comes from installation or configuration, and the report language follows that interface language unless explicitly set --language、OPENBUSINESS_OUTPUT_LANGUAGE or openbusiness config --language. The report is written to output/notion_business_model.md.

Environment variables override local configuration and are suitable for CI, containers, and temporary provider switching. Interface language priority:openbusiness analyze --ui-language en > OPENBUSINESS_UI_LANGUAGE=en > ui_language = "en"in local configuration. Report language priority:openbusiness analyze --language en > OPENBUSINESS_OUTPUT_LANGUAGE=en > output_language = "en"in local configuration.

Use --depth deep to get a more serious research process:standard keeps evidence collection bounded to speed up runs;deep lets the evidence collector spend more tool rounds and use a broader search depth, for customer proof, hiring signals, ecosystem clues, competitor positioning, and negative evidence. All analyst nodes still use the same depth standard: the main conclusions must include mechanism, evidence quality, counterexamples, business implications, and validation data.

Each run preserves a portable report path and writes a reproducible artifact folder containing run metadata, evidence, sources, and stage outputs, making it easier to debug, compare, and improve.

Applicable Scenarios

OpenBusiness is useful when you need a structured preliminary business model research report before conducting in-depth due diligence. Relevant keywords listed in its FAQ include: business model research CLI, business model canvas generator, company research tool, competitive analysis, unit economics, LangGraph agent, evidence-tagged report.

Limitations and Caveats

  • About private companies: they can be analyzed, but private company reports usually contain more inferences and missing claims, because private companies do not publish complete financial data, churn rate, CAC, ARPU, or cohort data. OpenBusiness is designed to show these gaps rather than fill them with fabricated numbers.
  • Run cost: reports cost about $0.10 to $0.40 in API calls, depending on the provider and evidence package size. The package is open source under the MIT License and requires bringing your own OpenAI, Anthropic, DeepSeek, Tavily, and Firecrawl keys.
  • Optional evidence APIs: if Tavily or Firecrawl is not configured, the pipeline will still run, but more claims will be marked as [INFERRED] or [MISSING].
  • Language purity: English reports will warn if the generated content contains Chinese characters; zh reports will also warn if the generated content contains English section headings or English-like paragraph lines. Company names, URLs, ticker symbols, metric abbreviations, API/tool names, model names, and evidence tags are intentionally preserved.
  • Custom analysis packages and templates are prompt guidance only: they cannot bypass evidence tags, nor can they turn missing data into verified claims.
  • Troubleshooting: issues such as./install.sh: Permission denied、openbusiness: command not found, reports being mostly inference, and mixed report languages are documented in the official documentation.

Project Information

  • Repository:https://github.com/wanikua/OpenBusiness
  • PyPI package: openbusiness 0.1.0
  • License: MIT License
  • Provides a static website GUI demo at site-demo/, showcasing the analysis cockpit, evidence tag counter, LangGraph pipeline, report preview, sources, and run artifacts.
  • Contributions: You can submit domain packs, report templates, evidence connectors, or better stage prompts; before submitting a PR, please read CONTRIBUTING.md, and refer to docs/EXTENDING.md when adding analysis packs, report templates, or evidence tools.

Related tools

Browse category