Skip to content

Connectors / Observability

Observability MCP connectors for Claude, ChatGPT and Cursor

10+ Observability connectors in the catalog, each behind the same governed MCP endpoint as everything else in Elaichi.

Start 14-day free trial Browse the full catalog
https://api.elaichi.ai/mcp The same for every user.

  • Their own access. An agent never gets more than the person it acts for.

  • One connection, every client. Claude, ChatGPT, Cursor, any MCP client, and the Elaichi Agent.

The connectors

Observability connectors in the catalog

Connect one and your whole team reaches it, each inside the access they already have, without anyone handling a credential.

11 connectors

Browse all 500+ connectors →

In practice

What teams do with Observability connectors

Drawn from the connector pages in this category, so every line describes something one of these connectors actually does.

Engineering

  • Datadog (AP1)

    Find the slow request after a deploy

    Ask which spans got slower in the last hour and have the agent set them beside the error logs for the same service.

  • Datadog (AP2)

    Trace a slow request end to end

    Fetch a trace by id, aggregate the spans by service and spot the hop that is taking the time.

  • Datadog (EU)

    Trace a slow request to the service at fault

    Search traces and spans for a request, aggregate them by service, and find where the time is going without clicking through flame graphs.

Support

  • Datadog (UK1)

    Check the logs behind a customer ticket

    Search the logs and real user monitoring events for the customer's session while the ticket is still open, so the reply to them is based on what actually happened.

  • Datadog (US1)

    Check whether a customer's problem is real

    When a customer reports errors, look up the RUM events and logs for their session in Datadog (US1) before replying, so the answer is grounded in what actually happened.

  • Datadog (US3)

    Check the logs behind a customer ticket

    Search logs for the customer's account around the time they reported the problem, and attach what you found to the ticket.

Platform

  • Datadog (US5)

    Spot hosts that stopped reporting

    Search hosts by tag or environment and ask which ones have gone quiet since the last deploy.

  • Grafana Cloud

    Find the dashboard nobody can locate

    Search dashboards and folders by name or keyword, get a summary of what each one shows, and send a deep link to the person who asked.

  • New Relic

    Check workload health before a release

    Ask which New Relic workloads are degraded right now and which services and hosts sit inside them, then decide whether to ship or hold.

On-call

  • Honeycomb

    Find the slow request behind a page

    Ask for the trace behind a latency alert, see which spans took the time, and get a plain summary before you open a laptop.

  • Datadog (AP1)

    Get the picture during an incident

    Pull up the incident in Datadog AP1, see which monitors fired, and read the logs and traces around it in one conversation instead of six tabs.

  • Datadog (AP2)

    Catch up on an incident at handover

    Get the current state of an open incident, the events around it and the spans that look slow, then write the handover note into a notebook.

Once it is connected

Things to ask

Each of these is answered against the access the person asking already has, in the Observability account you connected.

  • Which monitors fired for the checkout service in the last hour?

    Datadog (AP1)

  • Which monitors are alerting in production right now?

    Datadog (AP2)

  • Which monitors are firing for the checkout service right now?

    Datadog (EU)

  • Which monitors are alerting on the checkout service right now?

    Datadog (UK1)

  • Which monitors are alerting right now and on which hosts?

    Datadog (US1)

  • Which monitors are alerting in production right now?

    Datadog (US3)

  • Which monitors fired on the checkout service in the last hour?

    Datadog (US5)

  • Show the Prometheus error rate for checkout over the last hour

    Grafana Cloud

The tools

What an agent can call in Observability

Connectors
10+

in the Observability catalog

Tools
2,840

callable the moment you connect

Hosted sign-in
9

connect with nothing to register

Of those 2,840 tools, 5% delete something. Restricting an agent to reads is not a promise here, it is 1,496 tools admitted and the rest left out — and a restricted tool is never advertised to the model at all.

Read 53%
1,496 tools · list, get, search
Write 16%
446 tools · create, update, send
Delete 5%
143 tools · delete, remove, archive
Other 27%
755 tools · vendor-specific verbs

One endpoint

https://api.elaichi.ai/mcp

Every connector above answers here.

Every Observability connector, by depth

Tool counts are what the connector exposes today; the split is what those tools do.

Hosted sign-in takes a partnership with each vendor, and more are in progress. Until one lands, Your own app means the connector works today — you register an OAuth app once and connect. API key and App credentials mean an admin pastes them once, with nothing to register.

Observability connectors with tool counts split by read, write and delete
Connector Tools Read · write · delete Sign-in
Datadog (AP1) 352 Hosted
Datadog (AP2) 352 Hosted
Datadog (EU) 352 Hosted
Datadog (UK1) 352 Hosted
Datadog (US1) 352 Hosted
Datadog (US3) 352 Hosted
Datadog (US5) 352 Hosted
Vantage 226 API key
Grafana Cloud 62 Hosted
New Relic 54 API key
Honeycomb 34 Hosted

FAQ

Frequently asked questions

How many Observability connectors does Elaichi have?

10+ Observability connectors are in the catalog today, and the list grows as connectors are added. Each one arrives as a set of MCP tools an agent can call through https://api.elaichi.ai/mcp.

Can Claude, ChatGPT and Cursor all use Observability connectors?

Yes. Elaichi exposes one organization-wide endpoint, https://api.elaichi.ai/mcp, and Claude, ChatGPT, Cursor, any MCP client and the Elaichi Agent all connect to that same address with OAuth. Connecting a Observability account once makes it reachable from every one of them.

Do Observability connectors work with Gemini, Codex, Claude Code or other MCP clients?

Yes. Observability connectors are reached over the same MCP endpoint every client uses, so anything that speaks MCP can call them — Gemini, Codex, Claude Code, Windsurf, Cline, Zed and OpenCode among them — alongside Claude, ChatGPT, Cursor and the Elaichi Agent. There is no per-client setup beyond pointing the client at https://api.elaichi.ai/mcp.

Do Observability connectors need me to bring my own OAuth app?

No. 9 of the Observability connectors use Elaichi's hosted sign-in, with nothing to register. 2 connect with an API key or app credentials that an admin pastes once, with no app to register. Nothing in this category asks you to register an OAuth app before you start.

Can I stop an agent from writing to Observability tools?

Yes. Tool restrictions apply at role and individual level, and a restricted tool is never advertised to the model, so it cannot be called or guessed at from the tool list. Read-only access to a Observability connector is a matter of allowing the reads and leaving the writes out.

Whose access does an agent get on a shared Observability connection?

The access of the person the agent is acting for, resolved against their current role on every call — not the access of whoever connected the account. A colleague can use a connection without ever seeing its credential.

Put Observability connectors in front of your team

14 days on Gold, no credit card. Connect one and pick what each team can call.

Start free trial Talk to us
Works with
Claude ChatGPT Cursor and any other MCP client, or the Elaichi Agent.
When the trial ends
Nothing is deleted. Connections, roles and the audit log stay where they are, so subscribing picks up exactly where you left off.