Airbroke

Open Source Error Catcher

Features

> Airbroke is currently in active development! Using it in production environments is at your own discretion. We appreciate your feedback and support as we work towards a stable release.

• :floppy_disk: Based on PostgreSQL
• :globewithmeridians: Airbrake(tm)/Sentry(tm)-compatible HTTP collector endpoints
• :computer: Modern, React-based frontend for error management
• :rocket: Designed with simplicity at its core
• :wrench: Maintains small database footprint even under heavy data ingestion
• :robot: Ask AI about issues
• :clipboard: Provide cURL command to reproduce HTTP exceptions
• :arrow_forward: Replay HTTP exceptions
• :key: Supports multiple OAuth providers for secure user authentication
• :bar_chart: Occurrence charts
• :bookmark: Save and manage bookmarks for important occurrences
• :electric_plug: MCP API for LLM/agent integrations

System Requirements

• Node.js 22/24+ compatible environment
• Minimum of 300MB RAM
• At least 1000 millicores, equivalent to 1 CPU core
• PostgreSQL 15+ database
• 8+ free database connections slots per instance

Deployment Options

Airbroke provides flexibility in deployment options. You can either deploy it from the built source code or use a multiarch Docker image. For Kubernetes deployments, a Helm chart is provided. As Airbroke is a Next.js 16 application, it can be deployed wherever a Node.js server is supported. This includes managed environments such as Vercel, Netlify, and Heroku.

Build from source

For a production build, you can run:

cp .env.dist .env

Then edit the .env file to set your own values.

yarn install
yarn build

This will generate a .next build output. Because this repo uses output: "standalone", the runnable server output is in .next/standalone (see the Dockerfile for a working copy strategy / asset layout).

You can also run yarn start to test the production build locally on port 3000.

Docker

We publish images for both amd64 and arm64 architectures on ghcr.io but in case you want to build your own image you can do so.

You can build the Docker image with:

docker build --no-cache -t icoretech/airbroke:latest .

You can then run the image locally with:

docker run -p 3000:3000 icoretech/airbroke:latest

Vercel

[](https://vercel.com/new/clone?repository-url=https%3A%2F%2Fgithub.com%2Ficoretech%2Fairbroke&project-name=airbroke&repository-name=airbroke&env=DATABASEURL%2CDIRECTURL%2CBETTERAUTHSECRET%2CAIRBROKECORSORIGINS&envDescription=Required+env+vars+for+Airbroke+%28Prisma+%2B+Better+Auth%29.+Generate+a+random+BETTERAUTHSECRET.+Set+AIRBROKECORSORIGINS+to+your+app+origin%28s%29.&envLink=https%3A%2F%2Fgithub.com%2Ficoretech%2Fairbroke%2Fblob%2Fmain%2F.env.dist&envDefaults=%7B%22AIRBROKECORSORIGINS%22%3A%22*%22%7D&build-command=if+%5B+%22%24VERCEL_ENV%22+%3D+%22production%22+%5D%3B+then+corepack+yarn+db%3Amigrate%3B+fi+%26%26+corepack+yarn+build)

While testing on Vercel has not been conducted, Airbroke should be fully compatible.

It's important to keep the following points in mind:

• For optimal performance, ensure your database is located in the same region.
• The endpoints under /api/* will be converted into serverless functions, which may introduce potential cold boot time.
• If you connect directly to Postgres, use a pooler (or Prisma Accelerate / Data Proxy) to avoid exhausting DB connections.
• Migrations must be executed as part of deployment. The Deploy Button above sets a build-command that runs DB migrations only for VERCEL_ENV=production before building (adjust this in your Vercel project settings if you need a different workflow).
• The Deploy Button defaults AIRBROKE_CORS_ORIGINS to * for quick starts. After your first deploy, set it to your deployed origin (for example https://.vercel.app) and any custom domains.

Detailed instructions for this process can also be found in the Prisma deployment guide for Vercel.

Render.com

[](https://render.com/deploy?repo=https://github.com/icoretech/airbroke)

The Render Deploy Button uses render.yaml (Render Blueprint). On the free tier, the template runs yarn db:migrate at boot (safe no-op if already applied). Airbroke allows all origins if AIRBROKE_CORS_ORIGINS is unset — once Render assigns your public URL, lock it down in the Render dashboard for production (this controls CORS for /api/v3/notices and /api/sentry/*/envelope).

Railway

Railway supports one-click deployments via Templates. Because templates are created and published from within Railway, you’ll need to create a template once and then wire up the button URL.

[](https://railway.com/new/template/REPLACEME?utmmedium=integration&utmsource=button&utmcampaign=airbroke)

Steps:

• Create a Railway project from this repo (GitHub deploy).
• Add a PostgreSQL database in Railway.
• Set required environment variables (Railway can suggest vars from .env.dist).
• Publish the project as a Template and replace REPLACE_ME above with the template ID.

Netlify

[](https://app.netlify.com/start/deploy?repository=https://github.com/icoretech/airbroke)

The Deploy to Netlify button is configured via netlify.toml. The build command runs DB migrations only for production deploys.

Because you don't know your final *.netlify.app URL before the first deploy, you can start with AIRBROKE_CORS_ORIGINS=* and then tighten it to your deployed origin (for example https://.netlify.app) and any custom domains.

Heroku

[](https://www.heroku.com/deploy?template=https://github.com/icoretech/airbroke)

The Deploy to Heroku button uses app.json to provision add-ons and prompt for required environment variables.

Because you don't know your final *.herokuapp.com URL before the app is created, you can start with AIRBROKE_CORS_ORIGINS=* and then tighten it to your deployed origin (for example https://.herokuapp.com) and any custom domains.

Helm

You can deploy Airbroke to Kubernetes using the dedicated Helm chart.

The Helm chart includes a values.yaml file with some default values that you can override with your own. It also includes a pgBouncer chart as optional dependency.

When using Helm we recommend using a GitOps approach to deploy your application(s), such as Flux.

Please find more information about the Helm chart in the dedicated repository: icoretech/charts

Docker Compose

For users who prefer Docker Compose for managing multi-container Docker applications, a docker-compose.yml file is provided at the root of the repository.

To get started, make sure you have Docker and Docker Compose installed on your system. Then, you can start the application using the following command in the terminal:

make build
make run

You can override the default values in the docker-compose.yml file by creating a docker-compose.override.yml file in the same directory. This file is ignored by Git and will not be committed to the repository.

Setup

Please view all the available configuration variables in the .env.dist file.

Airbroke requires DATABASE_URL at runtime. DIRECT_URL is required when you need a direct database connection for migrations (for example, when DATABASE_URL goes through PgBouncer or a data proxy). Some examples:

# Example Connection to Prisma Data Proxy
DATABASE_URL="prisma://__HOST__/?api_key=__KEY__"
# Example Connection to PostgreSQL
DATABASE_URL="postgresql://__USER__:__PASSWORD__@__HOST__:__PORT__/__DATABASE__?connection_limit=20&pool_timeout=10&application_name=airbroke"
# Direct connection to the database, used for migrations
DIRECT_URL="postgresql://__USER__:__PASSWORD__@__HOST__:__PORT__/__DATABASE__"

The optimal connection pool size without pgBouncer (connection_limit) can be calculated using the following formula:

connection_limit = (num_physical_cpus * 2 + 1) ÷ number_of_application_instances

For a system with 8 CPU cores and 3 application instances, the calculation would proceed as follows:

connection_limit = (8 * 2 + 1) ÷ 3
connection_limit = (16 + 1) ÷ 3
connection_limit = 17 ÷ 3
connection_limit ≈ 5.67

Since connection_limit must be an integer, it should be rounded down to the nearest whole number. In this scenario, each of the 3 application instances should have a connection_limit of 5.

This limit can be set in your connection strings.

After deployment, you should be able to access your ingress (preferably secured with HTTPS) and start adding projects. This process will generate an API key that you can use with your Airbrake-compatible clients. This key, along with other essential information, will be provided to you.

About PgBouncer

To optimize your experience with Airbroke, as well as with Postgres overall, we advise integrating PgBouncer 1.24.0+ into your tech stack in transaction mode. For more comprehensive information, we recommend reviewing Prisma's Connection Management documentation, which provides insights on external connection poolers. We recommend not setting pgbouncer=true in the database connection string if you're using PgBouncer 1.21.0 or later (source).

About DIRECT_URL

The DIRECT_URL environment variable should be configured to establish a direct connection to the database. This is particularly crucial when using PgBouncer, as it enables migrations that cannot be executed through a data proxy. You can find more detailed information about this subject in the Prisma's guide on configuring pgBouncer.

Architecture

Frontend

The Airbroke frontend provides a user-friendly interface for managing and analyzing error reports. It utilizes a modern tech stack, including React, Tailwind CSS, and Next.js, to deliver a seamless user experience. The frontend leverages server-rendering capabilities to optimize initial page load times and ensure fast and responsive navigation.

To optimize performance, the frontend may implement caching strategies to reduce the number of database queries and enhance overall responsiveness. This ensures that you can efficiently navigate through error reports and analyze critical information without experiencing unnecessary delays.

Data Collection API

The Data Collection API is a core component of Airbroke responsible for handling the ingestion of error reports. It serves as the endpoint where clients can send error reports, enabling efficient data collection for error management.

To ensure simplicity and performance, the Data Collection API sidesteps the use of queue systems and performs parsing and transactions in-band. This means that parsing and processing of error reports happen synchronously within the API request cycle. Despite this approach, the Data Collection API demonstrates robust request-per-minute (RPM) performance even under high traffic volumes.

MCP API (LLM/Agent Integrations)

Airbroke exposes an MCP-compatible JSON-RPC API for read-only error triage workflows:

• POST /api/mcp

Authentication is static-header based (no OAuth in v1). Set AIRBROKE_MCP_API_KEY and send either:

• Authorization: Bearer
• X-Airbroke-Mcp-Key:

Origin policy for browser clients:

• Optional AIRBROKE_MCP_ALLOWED_ORIGINS (comma-separated) explicitly allows specific Origin values.
• If AIRBROKE_MCP_ALLOWED_ORIGINS is unset, Airbroke allows same-origin browser requests by default.
• Non-browser clients that do not send an Origin header are unaffected by this check.

Available tools:

• airbroke_list_projects
• airbroke_get_project
• airbroke_list_notices
• airbroke_list_occurrences
• airbroke_get_notice
• airbroke_search
• airbroke_get_occurrence
• airbroke_get_setup_guide

Notes on richer MCP outputs:

• airbroke_list_projects, airbroke_list_notices, and airbroke_list_occurrences

support offset + limit for page-like iteration.

• airbroke_list_projects is the recommended discovery-first tool for fresh conversations.

It matches both project names and organization names case-insensitively, so queries like my-app, school-portal, or example-org can surface the right project candidates without already knowing the internal project id.

• airbroke_get_project, airbroke_list_notices, airbroke_search, and

airbroke_get_setup_guide accept exact project ids, but also resolve common LLM-style project references such as partial project names when the match is unambiguous. If a reference is ambiguous, they return candidate projects instead of a silent false negative.

• airbroke_list_notices supports include_project=true to embed minimal project data

and include_resolved=false to hide resolved notices (default: true, show all). Notice payloads include resolved_at (null when unresolved).

• airbroke_list_occurrences is summary-first by default, and supports:
• include_details=true (+ optional backtrace_frames) for backtrace preview and key context fields.
• include_notice=true / include_project=true to embed parent notice/project context inline.
• include_resolved=false to hide resolved occurrences (default: true).
• When notice data is included, resolved_at is present in the notice payload.
• airbroke_get_notice returns:
• the notice payload (includes resolved_at)
• latest_occurrences (by updated_at desc)
• top_occurrences (by seen_count desc, then updated_at desc)
• optional occurrence detail snippets for faster LLM triage.
• include_resolved=false filters occurrences to unresolved only (default: true).
• Includes remarks_count for notice-level remarks.
• airbroke_get_occurrence returns full occurrence details with parent notice

(including resolved_at) and project. Includes remarks array with notice-level and occurrence-level remarks (id, body, username, createdat, isnoticelevel).

• airbroke_search supports cross-project query by occurrence message + notice/project metadata

with optional filters: organization, project_id, env, include_resolved, and rich snippets via include_details + backtrace_frames. Notice data in search results includes resolved_at.

• airbroke_get_setup_guide returns integration code snippets for connecting applications

to Airbroke. Supports Airbrake-compatible SDKs and Sentry SDKs. Optional project_id returns ready-to-use code with real credentials; without it, templates contain placeholders. Filter by sdk (airbrake or sentry) and/or framework (e.g. ruby, node, python). Each snippet explicitly disables non-error SDK features (APM, tracing, remote config, etc.) since Airbroke is an error-only collector.

Example Codex MCP server config:

[mcp_servers.airbroke]
url = "https://myairbroke.xyz/api/mcp"

[mcp_servers.airbroke.http_headers]
X-Airbroke-Mcp-Key = "replace-me"

Example discovery flow:

1. Call airbroke_list_projects with the best human-friendly hint you have, such as

search: "my-app" or organization: "example-org"

1. If needed, call airbroke_get_project with the returned id or the same unambiguous

project reference

1. Use the resolved project id for airbroke_list_notices / airbroke_search

Authentication Layer

Airbroke i

…

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: icoretech
• Source: icoretech/airbroke
• License: MIT
• Homepage: https://airbroke.icorete.ch

Install and usage instructions live in the source repository linked above.