# obsei

obsei redacts personal data from customer feedback before an LLM processes it, then finds themes and gives agents cited evidence. Run it on your own infrastructure.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 1441
- Author: obsei
- License: Apache-2.0 · policy=link-only · description only; the file stays with the author
- Source: https://github.com/obsei/obsei/tree/HEAD/packages/obsei
- Updated: 2026-10-21
- HTML: https://affpapa.org/en/nevibecode/mcp/obsei

## What it does

obsei processes customer feedback, removing personal data before an LLM sees it. It finds recurring themes and lets an AI agent answer questions with cited evidence. The service is self-hosted.

Use `obsei add` to configure sources and sinks, then run `obsei check` before `obsei run`. The `obsei init` command stores the database key and pseudonym salt in `.obsei/secrets.env`, a Git-ignored file accessible only to its owner. For servers and CI, set `OBSEI_DB_KEY` and `OBSEI_PSEUDONYM_SALT` or mount that file. Claude Code setup needs a database key; without it, the MCP server cannot be configured as shown.

## Requirements
- Needs: Python and uv (commands run via uvx)
- Environment variables: `OBSEI_CONFIG`, `OBSEI_DB`, `OBSEI_DB_KEY` — keys and tokens: ask the human for the values, never commit them or put them in shared files.

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add --env OBSEI_CONFIG=<OBSEI_CONFIG> --env OBSEI_DB=<OBSEI_DB> --env OBSEI_DB_KEY=<OBSEI_DB_KEY> --transport stdio obsei -- uvx obsei
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from obsei <command>

### Codex

**Platform CLI**
```bash
codex mcp add obsei --env OBSEI_CONFIG=<OBSEI_CONFIG> --env OBSEI_DB=<OBSEI_DB> --env OBSEI_DB_KEY=<OBSEI_DB_KEY> -- uvx obsei
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from obsei <command>

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "obsei": {
            "command": "uvx",
            "args": [
                "obsei"
            ],
            "env": {
                "OBSEI_CONFIG": "<OBSEI_CONFIG>",
                "OBSEI_DB": "<OBSEI_DB>",
                "OBSEI_DB_KEY": "<OBSEI_DB_KEY>"
            }
        }
    }
}
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from obsei <command>

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "obsei": {
            "type": "stdio",
            "command": "uvx",
            "args": [
                "obsei"
            ],
            "env": {
                "OBSEI_CONFIG": "<OBSEI_CONFIG>",
                "OBSEI_DB": "<OBSEI_DB>",
                "OBSEI_DB_KEY": "<OBSEI_DB_KEY>"
            }
        }
    }
}
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from obsei <command>

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add -e OBSEI_CONFIG=<OBSEI_CONFIG> -e OBSEI_DB=<OBSEI_DB> -e OBSEI_DB_KEY=<OBSEI_DB_KEY> obsei uvx obsei
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from obsei <command>

## When to use it
- redact personal data from feedback before LLM processing
- surface recurring themes in customer feedback
- give agents answers backed by cited feedback
- configure feedback sources and sinks

## Example prompts
- Find recurring themes in customer feedback and cite evidence for each one.
- Look for repeated complaints in the feedback and support the findings with citations.
- Answer my question using connected feedback sources and cite the evidence.

## Similar
- [content-modeling-best-practices](https://affpapa.org/en/nevibecode/skills/content-modeling-best-practices.md) — skill · 4800 installs
- [portable-text-conversion](https://affpapa.org/en/nevibecode/skills/portable-text-conversion.md) — skill · 2700 installs
- [analytics-tracking](https://affpapa.org/en/nevibecode/skills/kostja94-analytics-tracking.md) — skill · 2400 installs
- [performance-analyzer-sms](https://affpapa.org/en/nevibecode/skills/performance-analyzer-sms.md) — skill · 1700 installs
- [amazon-brand-analytics](https://affpapa.org/en/nevibecode/skills/amazon-brand-analytics.md) — skill · 1200 installs
- [traffic-analysis](https://affpapa.org/en/nevibecode/skills/traffic-analysis.md) — skill · 1200 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md