Your AI coding assistant already knows your codebase. AppSignal's hosted MCP endpoint and CLI bring it your production data too: errors, performance, logs, metrics, and more, without leaving your editor or shell.
MCP: bring your agent production data
Previously, connecting an AI agent to AppSignal meant running our MCP server as a Docker container. It worked, but it added setup overhead, especially if you wanted to try it out quickly.
That is no longer the default path. AppSignal MCP is a public HTTP endpoint at https://appsignal.com/api/mcp. Point your agent at it, sign in, and you are done. No containers, no infrastructure to maintain, no local process to keep running.
The Docker image remains available for environments that restrict outbound traffic and need a local proxy, but for most setups the public endpoint is all you need.
Where it works
AppSignal MCP works with coding agents and AI-powered editors — any MCP-compatible client. Connect from the agent you already use, or from the editor where you write code.
OAuth: sign in with your AppSignal account once. Access is granted at the application level and exposes all read and write tools at once. The simplest path for editors with native remote MCP support (Claude Code, VS Code) or that can run mcp-remote as a bridge (Cursor, Windsurf, Zed).
Bearer token: generate a long-lived MCP token under Account Settings, with fine-grained permissions per toolset. Set each area to read, write, or disabled, scope it to specific applications, and choose whether it auto-exposes new tools as they ship.
If you're just getting started, OAuth is the quicker path. If you need per-tool permissions or per-app scoping, generate a token instead.
What your agent can do
Your agent can read and update:
Error incidents: list, inspect, triage, assign, and add notes
Performance: slow actions, traces, and span trees
Anomaly detection: browse alerts and manage triggers
Logging: query log lines and configure ingestion rules
Metrics and dashboards: pull timeseries data and manage charts
App discovery: list apps, environments, namespaces, and deploy markers
Ask what broke, what slowed down, or what changed after a deploy. Natural-language questions replace tab-hopping through incidents, traces, and log views.
A few things are intentionally left out of the toolset, like notifier or user management, since those need owner-level permissions and are safer handled explicitly in the AppSignal UI.
The AppSignal CLI is a single Rust binary, nothing else to install, that brings the same monitoring data into your terminal. Tail logs as they arrive, search and update incidents, manage anomaly detection triggers, and work with dashboards and log-based metrics.
Or with the install script (macOS and Linux, Intel and ARM, including musl-based distros like Alpine):
Shell
curl -sSL https://github.com/appsignal/appsignal-cli/releases/latest/download/install.sh | sudo sh
Check it's on your path and see your auth status:
Shell
appsignal-cli --versionappsignal-cli about
Then sign in:
Shell
appsignal-cli auth login
Every command returns human-readable output by default. Pass --output json (or -o json) when your agent needs structured data instead, status messages still go to stderr, so a JSON stream stays valid JSON you can pipe elsewhere. Set up a project-local config so commands in a given repo default to the right organization without repeating flags.
The CLI also bundles an agent skill. Run appsignal-cli skill install to teach Claude, Codex, or OpenCode how to call AppSignal without custom glue code.
Errors, performance, logs, metrics, and dashboards. One AppSignal account, no duplicate setup between MCP and the CLI. Both are read/write gateways to data AppSignal already has, neither collects anything new on its own.
During an incident, pick the tool that matches where you already are: ask your agent in the editor, or run a command in the shell.
Start debugging, today
Our full monitoring toolkit is available on all plans, including our free plan.
