CoIsland is a native macOS monitoring app that brings alerts from code, deploys, data tools, tickets, mail, and on-call services into the Mac's notch. Engineers can schedule tool-specific queries, review matching items, and hand an alert's context to Claude Code, Codex, or Cortex Code while retaining approval control.
CoIsland is a native macOS monitoring app for engineers who need to follow activity across code, deployments, data systems, tickets, email, and on-call tools. It displays monitored activity in the Mac's notch rather than requiring a browser dashboard or a collection of open service tabs.
A monitor runs a query, search, filter, or threshold check against one connected tool on a schedule selected by the user. When a check produces a new match or crosses a configured threshold, CoIsland surfaces the alert with its matching rows or items, monitor context, and activity. Users can acknowledge, annotate, and resolve alerts from the app.
CoIsland also provides a one-click handoff to Claude Code, Codex, or Cortex Code. The agent receives the alert facts, query, retained results, and activity; it can investigate and request approval before acting, while the user keeps control over the alert's final resolution. Connections are made directly from the Mac, with credentials stored in the macOS Keychain.
Create monitors that ask one connected service a query, search, filter, or threshold question. Check intervals range from 15 minutes by default up to 30 days, and examples are available for tools such as GitHub, Jira, and Snowflake.
The Mac's notch shows one ring per monitor, the time until the next check, and the number of matching rows or items. New matches are distinguished from existing results so the current attention queue stays visible.
Open an alert to review its matching items, monitor query, connector, and activity. Users can acknowledge alerts, add notes, and resolve them manually; CoIsland does not close alerts on the user's behalf.
A button in an alert opens Claude Code, Codex, or Cortex Code in a terminal. The handoff includes the alert's facts, monitor query, retained rows, and activity, and the agent asks before actions that require approval.
CoIsland communicates with connected tools directly from the Mac rather than through a CoIsland server or account. Tokens are stored in the macOS Keychain, and Snowflake CLI connections using programmatic access tokens can be imported from connections.toml.
Monitors are saved as .sql files in ~/.coisland/watches and can be edited by hand. The setup flow supports testing a monitor before saving it, configuring alert conditions, and selecting an optional sound.
An engineer can monitor GitHub Actions, Vercel, GitLab, Cloudflare, Render, or similar deployment sources for failed runs and receive the relevant build or production item in the notch.
GitHub review requests and Jira or Linear queries can surface work assigned to the user, helping them see new reviews, tickets, or sprint-related changes without repeatedly checking each service.
A Snowflake monitor can run a SELECT, WITH, or SHOW statement such as a failed-task query and alert when new matching task records appear. Other listed data connectors include Databricks, dbt, PostHog, Supabase, PostgreSQL, MySQL, and ClickHouse.
When a monitored error, deployment, or data failure appears, the user can hand the alert's context to Claude Code, Codex, or Cortex Code for investigation while retaining approval over actions and final alert resolution.
The notch can consolidate monitors from different work contexts, such as a job, side project, or coursework, with each alert retaining its source and connector context.
CoIsland runs on Macs with Apple Silicon, including M1, M2, M3, M4, or later, running macOS 14 Sonoma or later. It does not run on Intel Macs.
The site lists 35+ connectors, including GitHub, Vercel, GitLab, Jira, Linear, Sentry, PagerDuty, Datadog, Snowflake, Databricks, Gmail, Calendar, HTTP/JSON, RSS/Atom, status pages, and shell commands. The complete capabilities depend on the connector and its available monitor types.
A monitor runs its configured query or filter on a schedule. It can alert when new rows or items appear, or when a count crosses a threshold such as count greater than zero. Existing matches on the first check do not count as new for row-based alerts.
CoIsland can hand an alert to Claude Code, Codex, or Cortex Code in a terminal. The agent receives the alert context and asks before actions that require approval. CoIsland does not close the alert automatically; the user resolves it.
Yes. Without a license, the two oldest monitors run with no stated time limit. The purchase page describes a one-time license for up to two Macs, a year of updates, and a 30-day refund period.
middleware.io
Full-stack observability for logs, metrics, traces, apps, infrastructure, and user experience, with OpsAI for AI-powered issue detection and remediation.
canyontechs.ai
CanyonTechs AI is an LLM-powered auto-remediation product that parses production logs, detects incidents, and delivers fixes as signed pull requests. It offers a workspace-based service with sign-up and login for individual and company users.
mezmo.com
AI-driven telemetry platform for filtering, routing, and analyzing logs, metrics, and traces
drdroid.io
AI SRE agent connecting tools for incident response and remediation
www.coldtea.ai
Coldtea is an AI-powered software delivery environment for teams that want to manage QA, production monitoring, bug detection, task creation, and fixes in one place. It runs on your machine to catch regressions and production issues after code is written.
rootly.com
AI-native incident management for engineering teams
