Obsei
Open source, privacy-first Voice of Customer analytics. A self-hosted alternative to Enterpret and Unwrap that strips PII from feedback before any LLM sees it.
Open source, privacy-first Voice of Customer analytics. A self-hosted alternative to Enterpret and Unwrap that strips PII from feedback before any LLM sees it.
This page is designed to help you find out whether Obsei is good and if it is the right choice for you.
Obsei is an open source, self-hosted Voice of Customer platform. It reads customer feedback from wherever it lives, removes personal data before any AI model sees it, and shows your team what customers are saying.
What it does
Who it's for
Product, support and CX teams that want what Enterpret or Unwrap offer but can't send customer text to another SaaS or a public LLM: banks, insurers, healthcare, public services, and companies working under GDPR or India's DPDP Act.
Obsei started in 2020 and was rewritten from scratch for 1.0 in 2026. It is Apache-2.0, with no paid tier, no hosted upsell and no telemetry. Install it with uv tool install "obsei[recommended]" or run the Docker image.
Listed in
PII redaction at ingest
Emails, phone numbers, card numbers, IBANs and national IDs for 16 countries are removed before any model, sink or agent sees the text. Checksums are validated where the format has one, such as Aadhaar and cards.
Self-hosted and air-gapped by default
Runs on a laptop, a server or Docker inside your network. The database is encrypted, there's no telemetry, and model and sink traffic to outside hosts is blocked until you allow it.
Local AI classification
Labels sentiment, intent and your own custom questions with a model that runs in process, with no API key needed. Ollama, vLLM, OpenAI, Azure, Bedrock and Vertex also work. Low-confidence answers go to a person for review.
Multilingual themes
Groups the same issue across 50+ languages into stable themes, with trend tracking, near-duplicate detection and k-anonymity, which hides groups that are too small.
20+ feedback sources
App Store, Google Play, Zendesk, Freshdesk, Intercom, Gong, GitHub issues, Hacker News, Bluesky, YouTube, RSS, email, SQL databases, CSV files, webhooks and any REST API or MCP tool.
Routing and alerts
Sends feedback to Slack, Microsoft Teams, Jira, Linear, GitHub issues or signed webhooks, using rules on labels and confidence. Alerts on theme spikes, new themes and pipeline health, plus regular digests.
MCP server for AI agents
Claude Code, Codex, Cursor, GitHub Copilot CLI and Gemini CLI can search feedback and get cited, redacted quotes. Any issue an agent proposes waits for human approval.
Privacy controls built in
Salted author pseudonyms, erasure on request (obsei forget), data export and an audit log. These support GDPR, India's DPDP Act, LGPD and CCPA work.
Guided setup, no YAML editing
obsei init creates a working project with sample data, obsei try <link> previews any app or repo, and obsei add slack sets up sources and sinks interactively.
Open Source
No paid tier, no open-core split and no hosted upsell. Signed container images, an OpenSSF Scorecard and a plugin system for your own connectors.
License
Apache-2.0
Obsei removes personal data before any AI model sees your customers' words. Emails, phone numbers, card numbers and national IDs are stripped as feedback arrives, authors become pseudonyms, and the database is encrypted. The default classifier runs on your own machine, so you can analyse reviews, tickets and calls without sending them to a SaaS vendor or a public LLM. It is also open source under Apache-2.0, with no paid tier, and it works with AI agents through MCP, so Claude, Cursor or Codex can answer "what are customers saying?" with real, redacted quotes.
Choose Obsei if you can't, or don't want to, send customer text to another company. Enterpret, Unwrap and Chattermill process your feedback on their servers and usually start at several thousand dollars a year. Obsei runs inside your own infrastructure, costs nothing, and lets you bring your own model, from a small local classifier to Ollama, Azure OpenAI or Bedrock. You also aren't locked in: the data stays in an encrypted DuckDB file you own, and every source, sink and model can be swapped.
Product, support and CX teams who want to understand customer feedback at scale but work under privacy rules: banks, insurers, healthcare, public services, and companies working under GDPR or India's DPDP Act. It also suits app developers and open source maintainers who want to track App Store, Google Play, GitHub or Hacker News feedback for free, and engineering teams building AI agents that need customer evidence without exposing personal data.
I started Obsei in 2020 as a small open source Python library to read app reviews and tweets with AI models. In 2022 it was a joint runner-up in the Reserve Bank of India's first global hackathon, HARBINGER, and my co-founder Girish Patel and I started a company, Oraika, around it. We had encouraging meetings with banks, but we were engineers with no experience of enterprise sales, and we couldn't close a deal. The project went quiet for almost three years.
In 2026 I came back to it. Customer feedback had become even more sensitive with LLMs everywhere, so I rewrote Obsei from scratch around one rule: understand customers without moving their personal data anywhere it shouldn't go. Version 1.0 shipped in October 2026. I maintain it in my free time, and the long-term plan is to move it to a neutral foundation such as LF AI & Data.
Python 3.11+ with Pydantic for configuration, Typer for the CLI, httpx for connectors and Starlette for the server. DuckDB with its encryption extension for storage. Hugging Face Transformers for the default zero-shot classifier (bge-m3-zeroshot-v2.0), fastembed for multilingual embeddings and GLiNER for optional name redaction. The official MCP Python SDK for the agent server. The Studio UI is plain TypeScript bundled with esbuild, and the docs and website are built with Astro.
We have collected here some useful links to help you find out if Obsei is good.
Check the traffic stats of Obsei on SimilarWeb. The key metrics to look for are: monthly visits, average visit duration, pages per visit, and traffic by country. Moreoever, check the traffic sources. For example "Direct" traffic is a good sign.
Check the "Domain Rating" of Obsei on Ahrefs. The domain rating is a measure of the strength of a website's backlink profile on a scale from 0 to 100. It shows the strength of Obsei's backlink profile compared to the other websites. In most cases a domain rating of 60+ is considered good and 70+ is considered very good.
Check the "Domain Authority" of Obsei on MOZ. A website's domain authority (DA) is a search engine ranking score that predicts how well a website will rank on search engine result pages (SERPs). It is based on a 100-point logarithmic scale, with higher scores corresponding to a greater likelihood of ranking. This is another useful metric to check if a website is good.
The latest comments about Obsei on Reddit. This can help you find out how popualr the product is and what people think about it.
Allow me to introduce Obsei (pronounced "Ob see" | /əb-'sē/). It is a low-code AI powered automation tool. Name was derived from three words OBServe, AnalyzE, Inform. - Source: dev.to / almost 5 years ago
Do you know an article comparing Obsei to other products?
Suggest a link to a post with product alternatives.
Is Obsei good? This is an informative page that will help you find out. Moreover, you can review and discuss Obsei here. The primary details have been verified within the last quarter. So they could be considered up to date. If you think we are missing something, please use the means on this page to comment or suggest changes. All reviews and comments are highly encouranged and appreciated as they help everyone in the community to make an informed choice. Please always be kind and objective when evaluating a product and sharing your opinion.
