sideseat/core repository overview

⁠SideSeat

⁠What is SideSeat?

AI agents are hard to debug. Requests fly by, context builds up, and when something fails you're left guessing.

SideSeat captures every LLM call, tool call, and agent decision, then displays them in a web UI as they happen. Run it locally during development, or deploy to your private cloud for team visibility.

Built on OpenTelemetry⁠ — the open standard already supported by most AI frameworks.

⁠Quick Start

npx sideseat

Open localhost:5388⁠ and instrument your agent. Here's Strands Agents⁠ as an example — see the setup guide⁠ for Vercel AI, Google ADK, LangGraph, CrewAI, AutoGen, OpenAI Agents, and others.

With SideSeat SDK — automatic setup, one import:

pip install strands-agents sideseat
# or
uv add strands-agents sideseat

from sideseat import SideSeat, Frameworks
from strands import Agent

SideSeat(framework=Frameworks.Strands)

agent = Agent()
response = agent("What is 2+2?")
print(response)

Without SideSeat SDK — manual OpenTelemetry setup:

export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:5388/otel/default
pip install 'strands-agents[otel]'

from strands.telemetry import StrandsTelemetry
from strands import Agent

telemetry = StrandsTelemetry()
telemetry.setup_otlp_exporter()

agent = Agent()
response = agent("What is 2+2?")
print(response)

SDKs: Python (PyPI)⁠ | TypeScript (npm)⁠

⁠Features

• Real-time tracing — Watch LLM requests and tool calls as they happen
• Message threading — See full conversations, tool calls, and images
• Cost tracking — Automatic token counting and cost calculation

⁠AI Agent Development with MCP

SideSeat includes a built-in MCP⁠ server that gives AI coding agents direct access to your agent's execution history — prompts sent, responses received, tool calls made, costs incurred, and errors encountered.

Connect your coding tool and let it optimize prompts, debug failures, and reduce costs using real observability data instead of guesswork.

# Kiro CLI
kiro-cli mcp add --name sideseat --url http://localhost:5388/api/v1/projects/default/mcp

# Claude Code
claude mcp add --transport http sideseat http://localhost:5388/api/v1/projects/default/mcp

# OpenAI Codex
codex mcp add --transport http sideseat http://localhost:5388/api/v1/projects/default/mcp

Config file for Kiro, Cursor, and other MCP clients:

{
  "mcpServers": {
    "sideseat": {
      "url": "http://localhost:5388/api/v1/projects/default/mcp"
    }
  }
}

See the MCP docs⁠ for all setup options.

Then ask your coding agent:

7 tools are available: list_traces, list_sessions, list_spans, get_trace, get_messages, get_raw_span, get_stats. See the MCP docs⁠ for setup guides for Kiro, Claude Code, Codex, Cursor, and other clients.

⁠Deployment

Local — Run on your machine with npx sideseat. Data never leaves your computer.

Self-hosted — Deploy to your private cloud for team-wide observability. See the configuration guide⁠.

⁠Compatibility

Agent frameworks — Strands Agents, LangGraph, CrewAI, AutoGen, Google ADK, OpenAI Agents, LangChain, PydanticAI

LLM providers — OpenAI, Anthropic, AWS Bedrock, Azure OpenAI, Google Vertex AI

Telemetry — Vercel AI SDK, OpenInference, MLflow, Logfire

⁠Resources

• Documentation⁠ — Setup, configuration, API reference
• Discussions⁠ — Questions and ideas
• Issues⁠ — Bug reports
• Contributing⁠ — Development guide

⁠License

AGPL-3.0⁠ — Free to use and modify. Distribute modified versions under the same license.