Skip to main content
Glama
adarshba

OpenObserve MCP Server

by adarshba
README.md
# OpenObserve MCP Server

[![npm version](https://badge.fury.io/js/openobserve-mcp.svg)](https://www.npmjs.com/package/openobserve-mcp)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![TypeScript](https://img.shields.io/badge/TypeScript-5.5-blue.svg)](https://www.typescriptlang.org/)

[Model Context Protocol](https://modelcontextprotocol.io) server for querying OpenObserve from AI agents (Claude, Cursor, OpenCode, etc.).

## Features

- **Multi-instance** — query multiple OpenObserve deployments in a single call
- **Adaptive strategy** — raw fetch ≤1 h, reservoir sampling 1–6 h, hourly aggregation 6 h–7 d, daily aggregation >7 d
- **Batch queries** — parallel multi-query execution with configurable concurrency
- **LRU cache** — TTL-based result caching with per-entry size cap
- **Cursor pagination** — stateless pagination across instances via base64 cursors
- **Langfuse tracing** — optional observability for every tool call

## Installation

```bash
npm install -g openobserve-mcp
# or without installing:
npx openobserve-mcp --config /path/to/config.json
```

## Configuration

Create a JSON config file:

```json
{
  "instances": [
    {
      "id": "prod",
      "name": "Production",
      "url": "https://openobserve.example.com",
      "auth": { "type": "env", "envVar": "PROD_O2_TOKEN" },
      "defaults": { "org": "default", "timeout": 30000, "maxResults": 1000 },
      "capabilities": ["logs", "traces", "metrics"],
      "tags": ["production"]
    }
  ],
  "batching": { "maxConcurrent": 5 },
  "caching": { "enabled": true, "ttl": 300, "maxSize": 1000 }
}
```

`auth.envVar` must point to an environment variable containing a Base64-encoded `user:password` string:

```bash
export PROD_O2_TOKEN=$(echo -n "user@example.com:password" | base64)
```

Config can also be passed inline via the `O2_MCP_CONFIG` environment variable (JSON string).

## MCP Client Setup

**Claude Desktop / Cursor** — add to your MCP config:

```json
{
  "mcpServers": {
    "openobserve": {
      "command": "npx",
      "args": ["openobserve-mcp", "--config", "/path/to/config.json"],
      "env": {
        "PROD_O2_TOKEN": "base64token"
      }
    }
  }
}
```

**OpenCode** — add to `.opencode/config.json`:

```json
{
  "mcp": {
    "openobserve": {
      "type": "local",
      "command": ["npx", "openobserve-mcp", "--config", "/path/to/config.json"],
      "environment": {
        "PROD_O2_TOKEN": "base64token"
      },
      "enabled": true
    }
  }
}
```

## Tools

| Tool | Description |
|------|-------------|
| `search_logs` | Search log streams with SQL across one or more instances. Automatically selects query strategy based on time range. |
| `batch_query` | Execute multiple SQL queries in parallel across instances. |
| `list_instances` | List configured instances with their capabilities and tags. |
| `list_streams` | List available streams on one or more instances. |
| `get_stream_schema` | Return field names and types for one or more streams (cached 10 min). |
| `get_logs_around` | Fetch log records surrounding a specific timestamp without writing SQL. |

### SQL Reference

```sql
-- Basic filter
SELECT * FROM "mystream" WHERE level = 'error'

-- Full-text search
SELECT * FROM "mystream" WHERE match_all('timeout*')

-- Aggregation
SELECT code, COUNT(*) FROM "mystream" GROUP BY code ORDER BY COUNT(*) DESC

-- Time bucketing (do not add _timestamp filters manually)
SELECT histogram(_timestamp, '1 hour') AS ts, COUNT(*) FROM "mystream" GROUP BY ts
```

Supported: `=`, `!=`, `>`, `<`, `>=`, `<=`, `IS NULL`, `IS NOT NULL`, `AND`, `OR`, `NOT`, `COUNT`, `SUM`, `AVG`, `MIN`, `MAX`, `GROUP BY`, `ORDER BY`, `match_all()`, `histogram()`. String literals use single quotes. Stream names use double quotes.

## Optional: Langfuse Tracing

Set the following environment variables to emit traces to Langfuse:

| Variable | Description |
|----------|-------------|
| `LANGFUSE_O2_ENABLED` | Set to `"true"` to enable |
| `LANGFUSE_O2_PUBLIC_KEY` | Langfuse project public key |
| `LANGFUSE_O2_SECRET_KEY` | Langfuse project secret key |
| `LANGFUSE_O2_BASE_URL` | Langfuse host (default: `https://cloud.langfuse.com`) |

## Requirements

- Node.js ≥ 18.0.0
- OpenObserve instance with API access

## License

MIT © [Adarsh BA](https://github.com/adarshba)

TDQS

A4.1/5.0

Scored across 6 tools

Disambiguation5/5

Each tool has a clearly distinct purpose: listing instances, searching logs, batch queries, listing streams, fetching schema, and retrieving context around events. No two tools overlap in functionality.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern (list_instances, search_logs, batch_query, list_streams, get_stream_schema, get_logs_around), with no mixing of conventions.

Tool Count5/5

Six tools is well-scoped for interacting with OpenObserve instances, covering instance discovery, log searching (single and batch), stream listing, schema inspection, and context retrieval without being overwhelming or sparse.

Completeness4/5

The tool surface covers the core workflows of exploring and querying log data, but lacks direct support for querying metrics or traces, which are capabilities mentioned in instances. Still, for log-focused operations, it is nearly complete.

Maintenance

ActivitySlowing
ResponsivenessNo issues