# Trodo

## Docs

- [Concepts](https://docs.trodo.ai/docs/agent-analytics/concepts.md): Core concepts and terminology in Trodo.
- [Anthropic](https://docs.trodo.ai/docs/agent-analytics/integrations/anthropic.md): messages.create — including streaming and tool use — is auto-captured as kind='llm' with full token extraction.
- [AWS Bedrock](https://docs.trodo.ai/docs/agent-analytics/integrations/bedrock.md): InvokeModel, Converse, and ConverseStream are auto-captured with per-provider model strings.
- [Cohere](https://docs.trodo.ai/docs/agent-analytics/integrations/cohere.md): chat, generate, embed, and rerank are auto-captured. Re-ranker fires as a retrieval span.
- [Google Gemini + Vertex AI](https://docs.trodo.ai/docs/agent-analytics/integrations/google.md): google-generativeai and vertexai auto-capture generate_content and streaming variants.
- [Raw HTTP / fetch](https://docs.trodo.ai/docs/agent-analytics/integrations/http-fetch.md): Outbound HTTP calls auto-capture as generic spans. Upgrade to trackLlmCall when tokens matter.
- [LangChain](https://docs.trodo.ai/docs/agent-analytics/integrations/langchain.md): Agents, chains, LCEL runnables, and @tool decorators all auto-capture as a nested span tree.
- [LlamaIndex](https://docs.trodo.ai/docs/agent-analytics/integrations/llama-index.md): Retrievers, query engines, and chat engines produce a span tree with retrieval + LLM children.
- [Mistral](https://docs.trodo.ai/docs/agent-analytics/integrations/mistral.md): Python-only auto-instrumentation for mistralai.chat and embeddings.
- [OpenAI](https://docs.trodo.ai/docs/agent-analytics/integrations/openai.md): chat.completions, responses, and embeddings are auto-captured as kind='llm' spans.
- [OpenTelemetry / OTLP](https://docs.trodo.ai/docs/agent-analytics/integrations/opentelemetry.md): Drop-in OTel ingest at /v1/traces — for NextJS + Vercel AI, Datadog/Jaeger coexistence, or any OTel-emitting framework.
- [Vercel AI SDK](https://docs.trodo.ai/docs/agent-analytics/integrations/vercel-ai-sdk.md): generateText, streamText, generateObject, and tool calls — requires experimental_telemetry opt-in.
- [Introduction](https://docs.trodo.ai/docs/agent-analytics/overview.md): Trodo is an observability platform for products and AI agents.
- [Quickstart](https://docs.trodo.ai/docs/agent-analytics/quickstart.md): Send your first trace to Trodo in under 2 minutes.
- [Tracing overview](https://docs.trodo.ai/docs/agent-analytics/tracing/overview.md): One wrapper, auto-captured spans, full waterfall in the dashboard.
- [Patterns](https://docs.trodo.ai/docs/agent-analytics/tracing/patterns.md): Helpers, trackLlmCall, feedback, custom attributes, errors, and tokens/cost — everything that sits on top of wrapAgent and withSpan.
- [Spans](https://docs.trodo.ai/docs/agent-analytics/tracing/spans.md): Nest any step inside a run. Kinds, fields, and which ones are auto-captured vs manual.
- [Spans outside wrapAgent](https://docs.trodo.ai/docs/agent-analytics/tracing/spans-external.md): How to emit spans from code that runs outside the wrapAgent callback — workers, queues, separate files, separate services.
- [startRun / endRun](https://docs.trodo.ai/docs/agent-analytics/tracing/start-run-end-run.md): Open a run in one process and finalise it in another — for sessions that span many HTTP requests over time.
- [trackMcp](https://docs.trodo.ai/docs/agent-analytics/tracing/track-mcp.md): Trace your own MCP server. One runless span per tools/call — no parent run, no session lifecycle to manage.
- [Troubleshooting](https://docs.trodo.ai/docs/agent-analytics/tracing/troubleshooting.md): Common problems and how to diagnose them.
- [wrapAgent](https://docs.trodo.ai/docs/agent-analytics/tracing/wrap-agent.md): Record one agent run. Everything inside the callback becomes part of it.
- [Boards](https://docs.trodo.ai/docs/analysis/boards.md): Create dashboards to organize and share your analytics
- [Breakdowns](https://docs.trodo.ai/docs/analysis/breakdowns.md): Segment metrics by properties for deeper analysis
- [Cohorts](https://docs.trodo.ai/docs/analysis/cohorts.md): Create and manage user segments for targeted analysis
- [Filters & Operators](https://docs.trodo.ai/docs/analysis/filters-and-operators.md): Complete reference for filter operators by data type
- [Flows](https://docs.trodo.ai/docs/analysis/reports/flows.md): Visualize the paths users take through your product
- [Funnels](https://docs.trodo.ai/docs/analysis/reports/funnels.md): Measure conversion through a series of steps
- [Insights](https://docs.trodo.ai/docs/analysis/reports/insights.md): Visualize trends and compositions in product behavior data
- [Reports Overview](https://docs.trodo.ai/docs/analysis/reports/overview.md): Insights, Funnels, Retention, and Flows for product behavior analytics
- [Retention](https://docs.trodo.ai/docs/analysis/reports/retention.md): Measure how often users return and engage with your product
- [Time Controls](https://docs.trodo.ai/docs/analysis/time-controls.md): Configure time ranges and granularity for your reports
- [Users](https://docs.trodo.ai/docs/analysis/users.md): Explore user profiles, activity, and segments
- [Changelog](https://docs.trodo.ai/docs/changelog/overview.md): Latest updates and releases
- [Client-side](https://docs.trodo.ai/docs/data-in/auto-events/client.md): Events automatically captured in the browser — no instrumentation required
- [Server-side](https://docs.trodo.ai/docs/data-in/auto-events/server.md): Uncaught exceptions from Node.js and Python backends, captured automatically
- [Implementation Guide](https://docs.trodo.ai/docs/data-in/implementation-guide.md): Patterns and best practices for implementing Trodo in AI and SaaS products
- [Installation](https://docs.trodo.ai/docs/data-in/installation.md): Install Trodo on any surface — web, Node.js, or Python
- [Naming Conventions](https://docs.trodo.ai/docs/data-in/naming-conventions.md): Best practices for naming events and properties in Trodo
- [Groups](https://docs.trodo.ai/docs/data-in/sdk-reference/groups.md): Associate users with organizations, teams, or any entity — from the browser, Node.js, or Python
- [Identify Users](https://docs.trodo.ai/docs/data-in/sdk-reference/identify.md): Link anonymous sessions to known user identities
- [User Profiles (People)](https://docs.trodo.ai/docs/data-in/sdk-reference/people.md): Set and manage persistent user properties from the browser, Node.js, or Python
- [Track Events](https://docs.trodo.ai/docs/data-in/sdk-reference/track.md): Send custom events from the browser, Node.js, or Python
- [Quickstart](https://docs.trodo.ai/docs/intro/quickstart.md): Get Trodo running on your app in a few minutes
- [What is Event Analytics?](https://docs.trodo.ai/docs/intro/what-is-trodo.md): Track what users do across your product — and turn that data into decisions.
- [What to Track](https://docs.trodo.ai/docs/intro/what-to-track.md): Plan your Trodo implementation for AI apps, agentic workflows, and modern SaaS
- [Data handling](https://docs.trodo.ai/docs/mcp/data-handling.md): What Trodo stores, what gets sent to the AI client, and how PII is gated.
- [Trodo MCP](https://docs.trodo.ai/docs/mcp/introduction.md): Connect Claude, Cursor, and any MCP-compatible client directly to your Trodo team's analytics and agent observability data.
- [Claude Code (CLI)](https://docs.trodo.ai/docs/mcp/quickstart-claude-code.md): Add Trodo to the Claude Code CLI over HTTP or stdio.
- [Claude Desktop](https://docs.trodo.ai/docs/mcp/quickstart-claude-desktop.md): Add Trodo as a custom MCP connector in Claude Desktop using OAuth.
- [Claude.ai (web)](https://docs.trodo.ai/docs/mcp/quickstart-claude-web.md): Connect Trodo to Claude.ai with OAuth in under a minute.
- [Cursor](https://docs.trodo.ai/docs/mcp/quickstart-cursor.md): Connect Trodo's MCP server to Cursor.
- [Custom client (Python / Node)](https://docs.trodo.ai/docs/mcp/quickstart-custom.md): Connect to Trodo's MCP server programmatically using the official MCP SDKs.
- [Rate limits](https://docs.trodo.ai/docs/mcp/rate-limits.md): How Trodo's MCP server caps request volume per token and per tool.
- [Scopes](https://docs.trodo.ai/docs/mcp/scopes.md): What each Trodo MCP scope grants, with the full tool-to-scope mapping.
- [Tool catalog](https://docs.trodo.ai/docs/mcp/tool-catalog.md): Every tool exposed via the Trodo MCP server, what it returns, and the scope it requires.
- [Troubleshooting](https://docs.trodo.ai/docs/mcp/troubleshooting.md): Common errors and how to fix them.
- [Agent Observability](https://docs.trodo.ai/docs/overview/agent-observability.md): Trace every agent execution end-to-end — LLM calls, tool invocations, latency, costs, and errors.
- [Event Analytics](https://docs.trodo.ai/docs/overview/event-analytics.md): Understand what users do across your product — acquisition, activation, retention, and everything in between.
- [Introduction](https://docs.trodo.ai/docs/overview/introduction.md): Trodo gives you complete visibility into your product and your AI agents — so you always know what to build next.
- [Basic agent](https://docs.trodo.ai/docs/recipes/basic-agent.md): The minimum viable Trodo-instrumented agent — one wrapAgent, one LLM call, one feedback signal.
- [Context manager (multi-file)](https://docs.trodo.ai/docs/recipes/context-manager.md): Spans from helper files attach to the active run automatically — no runId threading required.
- [Cross-service tracing](https://docs.trodo.ai/docs/recipes/cross-service.md): Carry run context across HTTP boundaries so a downstream service's work appears inside the caller's run.
- [Dual export](https://docs.trodo.ai/docs/recipes/dual-export.md): Send spans to Trodo and to your existing OpenTelemetry collector at the same time.
- [From scratch (raw fetch)](https://docs.trodo.ai/docs/recipes/from-scratch.md): No framework, no auto-instrument — raw HTTP to a custom LLM endpoint, fully traced via withSpan + trackLlmCall.
- [Multi-agent in-process](https://docs.trodo.ai/docs/recipes/multi-agent-in-process.md): How withSpan works across files inside a single wrapAgent — no runId passing required.
- [Multi-step agent](https://docs.trodo.ai/docs/recipes/multi-step-agent.md): Several stages in sequence — triage → research → write — each a named sub-step in the waterfall.
- [Streaming agent](https://docs.trodo.ai/docs/recipes/streaming-agent.md): Token-by-token streaming with full tracing — tokens and cost still roll up correctly.
- [Sub-agents](https://docs.trodo.ai/docs/recipes/sub-agents.md): parent_run_id vs joinRun — when a sub-agent should be its own run and when it should be a span.
- [Tool-calling agent loop](https://docs.trodo.ai/docs/recipes/tool-calling-agent.md): A Claude-style tool-use loop — N llm spans, N tool spans, conditional branches, and a retry after an error.

## OpenAPI Specs

- [openapi](https://docs.trodo.ai/docs/api-reference/openapi.json)