Sentry Agent Tracing

What it does

Sentry Agent Tracing takes in the execution of an AI agent run in your application: model calls, tool executions, MCP server interactions, and handoffs, along with HTTP, database, and queue spans. It produces connected traces showing inputs, outputs, token counts, cost, timing, and errors, plus dashboards for runs, error rates, model costs, and per-tool reliability.

People use it to find where an agent broke, such as a failing tool or slow database call that degrades output. Conversations replays a user session as a chat-like transcript that links each message to its underlying LLM and tool spans. Agent traces sit inside existing Sentry distributed traces, so they can be read alongside errors, logs, and session replays.

It suits developers running agents in Python or JavaScript/Node.js who already use Sentry or want application monitoring for AI features. Agent tracing is listed as available on every plan, including the free Developer plan, which is limited to one user. Setup requires installing the Sentry SDK and an integration for your agent framework. It uses the same Sentry account and usage allowances as the surrounding application monitoring.

What it can do

• Trace full agent runs from system prompt to final output as connected spans
• Replay user sessions as conversations linked to underlying LLM and tool spans
• Track model costs, token usage (input, output, cached, reasoning), and error rates in dashboards
• Monitor per-tool error rates, P95 latency, and invocation frequency
• Trace MCP tool calls as spans within agent traces
• Auto-instrument supported agent frameworks in Python and JavaScript without manual spans

Works with

• OpenAI Agents
• Vercel AI SDK
• LangChain
• LangGraph
• Pydantic AI
• Anthropic SDK
• Google GenAI
• Laravel AI
• MCP servers

Before you choose

• Some AI-specific features and quotas depend on plan; Seer AI debugging requires a subscription on the Team plan and above.
• Agent frameworks not auto-instrumented need manual instrumentation, which the FAQ estimates at about 10 lines per span type.
• Data retention varies by plan: 30-day lookback on Developer, up to 90-day on Business and Enterprise.

Listed under

• Coding
• DevOps and observability

Get started

Sign up, then install the SDK and enable tracing for your agent framework.

1. Create a Sentry account and project at sentry.io
2. Install the Sentry SDK for Python or JavaScript/Node.js and enable tracing with an agent framework integration
3. Name your agents and choose a sampling strategy in Explore → Agents

Agent tracing is included across Sentry plans. Team starts at $26/month and Business at $80/month when billed annually with default prepaid data; usage above allowances is billed separately.

Source: sentry.io/pricing, checked 30 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Agent Tracing: See Every Step of Your AI Agent Run | Sentry ↗sentry.io
• Agents | Sentry Docs ↗docs.sentry.io
• Pricing: Free Developer Plan, Pay as You Grow | Sentry ↗sentry.io
• Sentry for AI | Sentry Docs ↗docs.sentry.io

Company website

sentry.io: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.