FireHydrant MCP server: five read-only tools

FireHydrant exposes incident data through five tools. The context cost is higher than some nine-tool connectors, and a scope flag makes it read-only.

Last verified 1 September 2026 · from FireHydrant's repository, npm and the server's own tools/list response

This page is one of 90 in a directory of monitoring and developer tools. Each page checks what an AI coding agent can get out of the tool through its MCP server, the connector that lets the agent query the tool directly.

What is the FireHydrant MCP server?

It is FireHydrant's server for the Model Context Protocol. FireHydrant runs incident response, so what the agent gains is the incident record: what happened, who was alerted, and what the retrospective said.

The surface is narrow on purpose. Five tools cover listing incidents, listing alerts including Signals alerts, creating an incident, listing retrospectives for an incident, and updating a single retrospective field.

The server is generated from FireHydrant's API specification, which is why the tools map so directly onto endpoints and why the descriptions are short.

FireHydrant reaches part of the way into production observability through incident response, and reaches nothing in code review, release, agent observability or cost.

In practice, FireHydrant answers what happened during an incident and what the team concluded afterwards, and nothing about the system itself. An agent connected here can pull the incident record and the retrospective that followed. It cannot show you the failing query. Pair this connector with a telemetry connector.

What can an AI agent do with FireHydrant?

What connecting costs before the first question

A context window is the amount of text a model can hold at once, and connecting uses some of it before the agent asks anything. Every tool the server registers sends its name, description and input schema to the model when the session starts. The tool list is charged whether or not a single tool is called.

Five tools cost 5,858 tokens at the low end, more than OpenSearch's nine and more than Dynatrace's eighteen. Descriptions are short at 128 characters, so this is not verbosity.

The schemas carry it, at 3,241 characters per tool. A generated server describes its endpoints exactly, including every optional filter, and an agent receives all of that whether it filters or not.

Encodings: GPT-5.6, GPT-5.5 o200k_base; GPT-4, GPT-3.5 cl100k_base; Claude Haiku 4.5 to Opus 5 count_tokens. Counted 24 August 2026 from tools/list on the published firehydrant-mcp package. It was run locally with a placeholder key.

Registration is static, so no FireHydrant account was involved. Figures cover name, description and input schema.

The two models in column one share the o200k_base encoding, so they get one count. The spread is formatting, about 39%. Reasoning tokens are output, spent per request. The tool list is input, charged once at connection.

The Claude figures are measured rather than encoded. Anthropic publishes no local tokenizer. So the definitions were sent to the API's count_tokens endpoint, and a no-tools baseline was subtracted. The range covers two tokenizer generations rather than two serializations. Haiku 4.5 and Opus 4.6 sit at the low end. Opus 5 and Sonnet 5 sit at the high end. The gap between them is about a quarter.

FireHydrant MCP server limits: where answers come back incomplete

The server is generated from FireHydrant's API specification, and it has not been regenerated. Two of the three limits below follow from that.

How to configure FireHydrant MCP for agents

1. 1

   Run with --scope read unless writing is the point

   It is a genuine server-side gate, which most connectors in this guide do not offer.

2. 2

   Pin the version

   The package has not moved since 0.0.4 in February 2026, so pinning costs nothing and protects against a surprise publish.

3. 3

   Use --tool if five is still too many

   The flag mounts named tools only, which is a finer gate than scope.

4. 4

   Keep the transport local

   --transport sse opens a port; stdio does not.

Paste this into your agent’s instructions

An empty retrospective list and an incident with no retrospective look identical, and the generated server may simply lack a newer endpoint. The block makes both visible.

When you answer using FireHydrant MCP tools, state:
- The time range you searched, and how many incidents you
  retrieved against the limit you used.
- Whether a retrospective existed. An empty result means none
  was started, not that the incident had no cause.
- Whether the server is running with read or write scope.
Never state that data does not exist. State that none was
returned under the constraints above.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Often not. FireHydrant's REST API is documented and this server is generated from it, so a coding agent calling the API directly reaches strictly more than five tools.

The connector's advantage is the scope flag. Getting the same guarantee from raw API calls means trusting the agent, or building your own gate.

Given the package has not moved since February 2026, the API is also the surface that will keep pace with the product.

Connect the server with read scope for incident investigation. Use the API where you need an endpoint the generated server never got.

FireHydrant MCP server setup

The package runs with npx and takes a FireHydrant API key on the command line. The start subcommand is required; running the package alone prints usage.

--scope accepts read or write and is repeatable, which is how the tool list is gated at the server.

claude mcp add-json "firehydrant" '{"command":"npx",
  "args":["-y","firehydrant-mcp","start",
          "--scope","read","--api-key","<your-key>"]}'

{"mcpServers":{"firehydrant":{"command":"npx",
  "args":["-y","firehydrant-mcp","start","--scope","read",
          "--api-key","<your-key>"]}}}

[mcp_servers.firehydrant]
command = "npx"
args = ["-y", "firehydrant-mcp", "start", "--scope", "read", "--api-key", "<your-key>"]

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"mcpServers":{"firehydrant":{"command":"npx",
  "args":["-y","firehydrant-mcp","start","--scope","read","--api-key","<your-key>"]}}}

{"mcpServers":{"firehydrant":{"command":"npx",
  "args":["-y","firehydrant-mcp","start","--scope","read","--api-key","<your-key>"]}}}

{"servers":{"firehydrant":{"command":"npx",
  "args":["-y","firehydrant-mcp","start","--scope","read","--api-key","<your-key>"]}}}

{"mcpServers":{"firehydrant":{"command":"npx",
  "args":["-y","firehydrant-mcp","start","--scope","read","--api-key","<your-key>"]}}}

npx -y firehydrant-mcp start --api-key <your-key> \
  --tool list-incidents --tool list-retros-by-incident

Read-only access, permissions and security

• --scope read is a real server-side gate

  Write tools do not appear in the tool list at all, so an agent cannot call them however the client is configured.

• Declaring an incident pages people

  create-incident is not a draft; it is the same action a human takes to start an incident response.

• --tool narrows further

  Mounting only the tools you need is finer than scope, and useful where read access itself should be partial.

• The package is frozen at 0.0.4

  A generated server that is not regenerated drifts from its API, so treat missing capability as a packaging gap rather than a product one.

Troubleshooting

The server prints usage and exits

The start subcommand is required. npx firehydrant-mcp start --api-key <key> is the working form.

The agent has no tool for creating an incident

The server is running with --scope read, which removes the write tools from the list. That is the intended behavior for investigation.

An incident has no retrospective

None was started. list-retros-by-incident returns what exists, and an empty result is not evidence that the incident was uncomplicated.

An endpoint you use is not available as a tool

The server is generated from FireHydrant's API specification and has not been regenerated since February 2026. Call the REST API directly for anything newer.

FireHydrant MCP server: Reference

What engineers report

FireHydrant's connector has a very small public footprint, so this section reports counts rather than quotes.

Repository facts from the GitHub API and package facts from npm, both retrieved 24 August 2026. Tool counts and schema sizes come from the server's own tools/list response.

Should you connect an agent to FireHydrant?

Yes, though it is no longer actively maintained. Best for read-scope investigation, with the REST API kept in reserve for anything the five tools do not reach.

Strongest fit

FireHydrant teams who want an agent reading incident history and retrospectives during or after an outage.

Main advantage

A genuine server-side read-only scope, which most connectors in this guide do not offer.

Main weakness

Five tools is a narrow slice of FireHydrant, and the package has not been regenerated since February 2026.

Operational risk

Under write scope an agent can declare an incident, which pages people.

Frequently asked questions