Menu

Choose a theme and configure high-contrast mode. Preferences are saved in your browser only.

User preferences

Theme

Pick a palette or follow your system preference.

High contrast

Sharper text and borders. System follows your OS setting.

Self-hosted error tracking.

Error tracking.
Minus the baggage.

Find what broke in production and why. Deploy Bicho to your own Cloudflare account, connect your app, and get back to shipping. No servers to babysit.

Open source. No Bicho subscription. Just your Cloudflare bill.

CaptureOTLP / HTTP

Catch it at the source.

Send the error straight to your Cloudflare Worker over OpenTelemetry.

Track errors.
Not infrastructure.

Capture
Your app breaks. Bicho takes the report. Send errors with a Bicho SDK, an existing Sentry SDK, or an OpenTelemetry logs exporter over OTLP/HTTP.
Keep
Same bug again? Bicho groups repeated errors into issues and keeps the original events. A Worker handles ingestion. Your D1 and R2 hold the data.
Investigate
Bring your MCP client or use the API. Search issues, dig into stack traces, and mark an issue resolved without throwing away the evidence.

Already using Sentry?

Keep your SDK and error-capture code. Create a Bicho project, swap in its generated DSN, and send new errors to your own Cloudflare account.

Bicho handles error tracking. Disable tracing, replay, profiling, and other unsupported integrations. Your existing Sentry history stays where it is.

Move from Sentry

Starting fresh? Use a Bicho SDK for JavaScript / Node.js, PHP, Rust, or Python. Building with Astro or WordPress? Use the Astro integration or WordPress plugin. Other languages can use an OTLP logs exporter. Try the runnable SDK examples.

Your existing Sentry initialization
Sentry.init({
  dsn: 'https://YOUR_PUBLIC_KEY@collector.example/PROJECT_ID',
  tracesSampleRate: 0,
  sendClientReports: false,
});

Wall of love.

“I wanted to track exceptions. Not make an exception in the budget.”

Count von Stacktrace

“To understand the bug, read the stack. To understand the invoice, seek enlightenment.”

Master Stack Fu

“Self-hosted? I cleared my weekend. Now I have to make actual plans.”

Sir Deploys-a-Lot

“The bug has been with us since sprint twelve. We call it an interface now. Other teams depend on it.”

Captain Backlog

“My agent read the stack trace before I finished explaining why it couldn't be my code.”

Agent Null

“There is no secret ingredient. It's a null pointer.”

Grandmaster Null Pointer

* Imaginary fans. Genuine enthusiasm. We made up the reviews for a little fun.
The bugs wrote themselves.

Your tracker. Your data.

Deploy your own copy of Bicho on Cloudflare Workers.
It runs in your Cloudflare account. Your D1 and R2 keep the data.

Frequently asked questions.

Agentic-first error tracking. Your agent reads the evidence, so you can get on with the fix.

What is Bicho?

Bicho is an open-source, self-hosted error tracker built for AI agents. It captures production errors, groups repeated failures into issues, and gives your agent the evidence through Model Context Protocol (MCP). It runs in your own Cloudflare account. Bicho means "bug" in Spanish. Bicho.FYI: the bug, for your information.

How Bicho works

Why is Bicho agentic-first, with no dashboard?

Your agent is the interface. Bicho has no dashboard UI to learn or click through. Ask your LLM-based coding agent to find production errors, read the stack traces, and investigate while you stay in your development workflow. Spend your time fixing the bug instead of hunting through dashboards and copying error reports into a chat.

Investigate with MCP

What can an AI agent do through Bicho MCP?

A connected agent can search issues and events, inspect stack traces and captured context, compare errors across releases, and check processing health. It can also create projects, recover a project’s public ingest configuration, and mark issues resolved, ignored, muted, or unresolved. Approve state changes deliberately; captured error text is evidence, not instructions for your agent.

Available MCP tools

Which AI agents and MCP clients can connect?

Use a client that supports remote MCP over Streamable HTTP and a custom Authorization header. Connect it to your own Bicho instance's /mcp endpoint with your admin bearer token. Configuration differs by client, so check its remote-server and secret-storage settings.

Connect your agent

Does the public Bicho MCP have access to my errors?

No. The public MCP at bicho.fyi/mcp serves documentation and setup guidance only. It cannot read your errors or administer your installation. Your own instance has a separate authenticated MCP endpoint for investigation. Never send your admin token to the public documentation server.

Documentation vs. investigation MCP

Can I use Bicho without an LLM or AI agent?

Yes. MCP is the primary agent workflow, but Bicho also exposes an authenticated HTTP API. You can search errors and manage issue status from scripts or your own tools without connecting an LLM. Bicho does not include a built-in graphical interface.

HTTP API reference

Where is my error data stored?

Your installation runs on Cloudflare Workers. D1 stores searchable issue and event metadata, while a private R2 bucket holds captured evidence. Those resources belong to your Cloudflare account. You control access, backups, and storage; this documentation website does not host your application errors.

Storage and access

Is Bicho free and open source?

The Bicho core, JavaScript, PHP, Rust, and Python SDKs, Astro integration, and WordPress plugin are open source under Apache-2.0. There is no Bicho subscription. You pay for any Cloudflare resources your installation uses and any AI service you choose to connect. Hosting costs depend on your traffic, queries, and stored data.

Software licenses

How do I set up self-hosted error tracking with Bicho?

Deploy the server to a Cloudflare account with R2 enabled, save your admin token securely, and connect your MCP client. Create a project, then configure your application with the generated ingest endpoint and public key. Installing the SDK alone does not deploy the server.

Deployment and first-project setup

Which languages have Bicho SDKs?

Bicho has SDKs for JavaScript and TypeScript, including Node.js, plus PHP, Rust, and Python. The JavaScript SDK can automatically capture uncaught browser errors and unhandled promise rejections. Backend reporting is explicit: JavaScript queues events for flush or close, PHP and Python send synchronously, and Rust delivery is awaited on Tokio. The Rust SDK also works in Tauri applications. For Astro sites, the integration adds opt-in browser capture and server request-exception reporting. For WordPress sites, the plugin reports PHP warnings and fatal errors alongside frontend browser errors.

SDKs and integration guides

Can I keep my existing Sentry SDK?

Yes. Create a Bicho project and use its generated Sentry DSN in your existing SDK configuration. Keep your error-capture code and disable integrations for tracing, replay, profiling, and other unsupported data types. This redirects new errors; it does not import your Sentry history or reproduce Sentry grouping. Secret-enabled backend projects also need the generated transport headers.

Move from Sentry

Can other languages send errors through OpenTelemetry?

Bicho accepts OTLP/HTTP logs in JSON or binary protobuf. Other languages can use an OpenTelemetry logs exporter when their instrumentation emits exceptions or error-level records. Configuring an exporter does not install exception capture, and trace-only exceptions are not logs. You do not need a separate OpenTelemetry Collector.

OTLP integration guide

How does Bicho group repeated errors?

Bicho groups repeated failures into issues while keeping their individual events for investigation. Your agent can inspect an issue, then read the captured events behind it. Resolving an issue changes its status without deleting its evidence; new occurrences do not automatically reopen it.

Issues, events, and status

Does Bicho collect sessions, page views, or performance traces?

No. Bicho focuses on error tracking, not session replay, product analytics, metrics, or distributed tracing. The JavaScript SDK does not automatically collect cookies, browser storage, user identity, console output, or network activity. Messages, stacks, and custom context can still contain sensitive data supplied by your application.

What the SDKs collect

How should I protect sensitive data when using an AI agent?

Send only the evidence you intend administrators and connected tools to read. Bicho filters recognized credentials, but it cannot remove every secret or personal detail. A connected agent can pass retrieved evidence to its LLM provider, so review that provider and your client settings. The admin token grants broad access across projects, including writes; it is not a read-only credential.

Privacy and credential boundaries

Does Bicho automatically delete old errors?

No. Events are append-only, and the core has no automatic retention deletion or event-delete endpoint. Resolving, ignoring, or muting an issue does not remove its stored events. Plan for storage growth and manage backups and recovery for both D1 and R2.

Storage and operating limits

Is Bicho a replacement for a full observability platform?

Bicho fits teams that want self-hosted error tracking and MCP-based investigation without another dashboard. It is not a full observability suite: it has no built-in alerts, session replay, metrics, or trace ingestion. Keep separate tools for those jobs if your application needs them.

Scope and limitations

Still on the fence?

Fine. Keep watching errors pile up in the console.
Or just try it. It's your Cloudflare account, not ours.

Type to search the documentation.