Last month I released Mira, an open-source system for company data enrichment. It automates the process of gathering and structuring company information, something I found myself doing repeatedly for sales, investment, and market research workflows.
You start by defining an agent, a workflow that specifies the data points you want (for example funding, industry, or hiring signals), the sources to query (websites, LinkedIn, Google Search), and optional evaluation criteria (such as "raised Series A in the last 12 months").
Mira then runs specialized agents. Website Explore looks beyond the homepage and checks relevant subpages like /press or /careers. The LinkedIn agent pulls structured company data and posts. The Search agent generates targeted queries and extracts details from results. The Analysis agent applies your criteria and produces a fit score with reasoning.
The output is a structured profile with confidence scores and source attribution. If all data is found with high confidence, the run stops early to save compute.
Since release, I've added configurable agents, bulk processing of company lists, evaluation with fit scoring, and outreach drafting from research results. Mira has been used for market mapping, filtering startup lists into targeted shortlists, enriching CRMs, and creating personalized outreach.
It is MIT licensed, built with TypeScript and the OpenAI Agents SDK, and comes with a Next.js app to visualize the process step by step.
Last month I released Mira, an open-source system for company data enrichment. It automates the process of gathering and structuring company information, something I found myself doing repeatedly for sales, investment, and market research workflows.
GitHub: https://github.com/dimimikadze/mira
You start by defining an agent, a workflow that specifies the data points you want (for example funding, industry, or hiring signals), the sources to query (websites, LinkedIn, Google Search), and optional evaluation criteria (such as "raised Series A in the last 12 months").
Mira then runs specialized agents. Website Explore looks beyond the homepage and checks relevant subpages like /press or /careers. The LinkedIn agent pulls structured company data and posts. The Search agent generates targeted queries and extracts details from results. The Analysis agent applies your criteria and produces a fit score with reasoning.
The output is a structured profile with confidence scores and source attribution. If all data is found with high confidence, the run stops early to save compute.
Since release, I've added configurable agents, bulk processing of company lists, evaluation with fit scoring, and outreach drafting from research results. Mira has been used for market mapping, filtering startup lists into targeted shortlists, enriching CRMs, and creating personalized outreach.
It is MIT licensed, built with TypeScript and the OpenAI Agents SDK, and comes with a Next.js app to visualize the process step by step.
Demo video: https://www.youtube.com/watch?v=NPTLzECkBT8
Happy to answer questions or discuss the approach.