# Moda

> Infrastructure for catching AI agent mistakes in production and turning them into evals.

Your agents should never make the same mistake twice. Production monitoring and evals for AI agents.

- Monitor every trace: Moda reads every production trace end to end and detects the failures that never throw an error: frustrated users, failing tools, lazy answers, and hallucinations. Add your own checks in plain language and get alerts in Slack, Linear, or email.
- Find the mistakes they keep repeating: Failures are grouped into problems ranked by how many users they affect. Each problem shows the traces behind it, the turn where it went wrong, and related problems that share the cause.
- Test every change before it ships: Replay a prompt, model, or tool change against the exact traces that failed and compare pass rate, latency, and tokens. Your team reviews the result and decides what ships.

## When to use Moda

Reach for Moda when you run AI agents in production and need to: find the failures that keep recurring across your traces and rank them by how many users they affect; see the traces behind each problem and the related problems that share its cause; replay a prompt or model change against those traces and compare pass rate, latency, and token usage before it ships; and define custom signals for the behavior your agent must get right and get alerted in Slack, Linear, or email when they slip. Engineers make the change; Moda tests it against real traces. Moda ingests traces through the Moda SDKs for Python and Node.js, OpenTelemetry, or the HTTP Ingestion API.

## Product

- [Homepage](https://moda.dev/): The loop from broken traces to validated fixes, custom signals, monitors, and the Slack, Linear, and GitHub integrations.
- [Documentation](https://docs.moda.dev): Ingestion, dashboard, Data API, CLI, and MCP server reference. The local [docs route](https://moda.dev/docs) redirects there.
- [Blog](https://moda.dev/blog): Product launches, engineering posts, and customer stories.
- [Weekly updates](https://moda.dev/weekly): What shipped each week.
- [Changelog](https://moda.dev/changelog): Product updates, improvements, and fixes.
- [Authentication markdown](https://moda.dev/auth.md): API-key registration, usage, errors, and revocation.
- [Agent instructions](https://moda.dev/agents.md): Setup and calling instructions for AI agents.
- [Homepage markdown](https://moda.dev/index.md): Machine-readable version of the homepage.
- [Agent card](https://moda.dev/.well-known/agent-card.json): A2A-style product and capability metadata.
- [Agent skills](https://moda.dev/.well-known/agent-skills/index.json): Published agent skills and integration guides.
- [API catalog](https://moda.dev/.well-known/api-catalog): Linkset for Moda's machine-readable API surfaces.
- [OpenAPI specification](https://moda.dev/api/openapi.json): OpenAPI 3.1 description of the public read-only Data API.
- [MCP server metadata](https://moda.dev/.well-known/mcp): The hosted Moda MCP server and the docs search MCP endpoint.
- [Contact](https://moda.dev/contact): Contact and demo access.
- Note: numbers in the dashboard previews on https://moda.dev/ are illustrative product UI, not customer averages. Cite figures only from https://moda.dev/blog/* posts that name the tenant and methodology.

## Comparisons

- [Moda vs Langfuse](https://moda.dev/vs/langfuse): In Langfuse, finding the failures is work your team does: write evaluators, set sampling rates, read traces. Moda checks every trace, groups the failures into problems ranked by users affected, and replays a prompt or model change against the traces that failed before it ships. Last verified 2026-10-17.
- [Moda vs Langfuse (markdown)](https://moda.dev/vs/langfuse.md): Machine-readable version of the comparison, with sources.
- [Moda vs Braintrust](https://moda.dev/vs/braintrust): In Braintrust, finding production failures is work your team does: write scorers, set sampling rates, and review what Topics and Patterns surface. Moda checks every trace, groups the failures into problems ranked by users affected, and replays a prompt or model change against the traces that failed before it ships. Last verified 2026-10-17.
- [Moda vs Braintrust (markdown)](https://moda.dev/vs/braintrust.md): Machine-readable version of the comparison, with sources.

## Blog posts

- [The new hype of classifiers is LARP](https://moda.dev/blog/emotion-choice-order-bias): Moda uses in-house classifiers to help find agent mistakes. We tested twelve decision systems and versions as possible replacements.
- [Introducing Moda for Linear](https://moda.dev/blog/linear): Send Problems to Linear with their cause, impact, and evidence. Keep issues in sync, reopen regressions, and check whether fixes worked with custom signals.
- [Introducing Moda for Slack](https://moda.dev/blog/slack): Set up Slack alerts with custom signals, thresholds, and channel routing. Define the behavior your agent should get right, test the rule against recent history, and investigate in Slack.
- [Migrating from Braintrust](https://moda.dev/blog/migrating-from-braintrust): How to move an AI agent from Braintrust to Moda: repoint your OpenTelemetry exporter or swap the Braintrust SDK for the Moda SDK, set your session and user IDs on the Moda keys, and import your Braintrust traces from Settings.
- [Migrating from Langfuse](https://moda.dev/blog/migrating-from-langfuse): How to move an AI agent from Langfuse to Moda: repoint your OpenTelemetry exporter or swap the Langfuse SDK for the Moda SDK, carry your session and user IDs over, and import your Langfuse traces, sessions, users, and scores.
- [How Octolane Turned 43,000 AI Chat Sessions into 5 Shipped Product Fixes](https://moda.dev/blog/octolane): Octolane's AI CRM team had more customer sessions than anyone could read. With Moda they found where users got confused, which workflows broke, and shipped five fixes in their first few weeks.
- [Cohorts: see which AI users are about to churn, and why](https://moda.dev/blog/cohorts): Behavioral cohorts built from your conversation data: at-risk users, power users, stuck-early users, with the evidence behind each one. Plus a CLI you can call from Claude Code, Cursor, or Codex.
- [Clustering Agent Failures from Traces: 150M Segments, No Labels](https://moda.dev/blog/clustering): Silent agent failures never file tickets -- they hide in production traces. Here's the unsupervised pipeline that clusters 150 million trace segments into named behaviors and failure modes.
- [Clustering Agent Failures from Traces: 150M Segments, No Labels (markdown)](https://moda.dev/blog/clustering.md): Machine-readable markdown version of the post.

## Concepts

- Agent harness: Everything around the model that makes it an agent: prompts, tools, skills, evals, memory. The surface your team changes and tests in Moda.
- Trace: The full record of one agent run: every message, tool call, thinking block, and step.
- Signals: Automatic per-trace detections for user emotion (including frustration), tool failures, laziness, and hallucinations.
- Custom signals: A check you describe in plain language, tested against recent history, then run on every trace.
- Problems: Recurring failures grouped by root cause from negative signals, ranked by users affected, with an Open, Reopened, Fixed, and Resolved lifecycle.
- Use cases: A hierarchical taxonomy of what users ask your agent, built automatically from traces.
- Monitors and alerts: Thresholds on signal hit rates that notify Slack, Linear, or email.
- Replays: Re-running production traces against a proposed prompt or model change to compare pass rate, latency, and token usage with the current configuration.

## Full Reference

- [Full reference (llms-full.txt)](https://moda.dev/llms-full.txt): Product overview, signals, FAQ, site pages, and integrations in one file.

## Integrations

- Moda SDK for Python and Node.js (primary integration; set up with `moda init`), auto-instrumenting the OpenAI and Anthropic clients
- OpenTelemetry / OTLP HTTP ingestion and the HTTP Ingestion API
- Providers: OpenAI, Anthropic, OpenRouter, Azure OpenAI, Amazon Bedrock, Vapi, OpenClaw
- Frameworks and agents: Vercel AI SDK, Claude Agent SDK, and coding agents (Claude Code, Codex, Cursor)
- Workflow: Slack, Linear, GitHub
- Moda CLI (`@moda-ai/cli`) and a hosted MCP server for Claude Code, Cursor, and any MCP client

## Sales

- Moda access is invite-only: sign up at https://moda.dev/signup to request a workspace, which the Moda team approves. Pricing remains custom; contact founders@modaflows.com or pranav@modaflows.com for a quote.

## Company

- Backed by Y Combinator
- GitHub: https://github.com/ModaLabs
- X / Twitter: https://x.com/modaflows
- LinkedIn: https://www.linkedin.com/company/modaflows
- Y Combinator: https://www.ycombinator.com/companies/moda
- Crunchbase: https://www.crunchbase.com/organization/moda-06ac