Agents & MCP

Talk to your infrastructure. Keep the context.

Explore Kapten’s agent workflow: ask about an environment, investigate a service or follow a deployment from the coding tools you already use.

Request early accessExplore Kapten MCP
Kapten MCPWith Kapten, Navigator, Helmsman and 3 more
claude~/my-appkapten · connected
  2. kapten · analyze_repo(path: ".")Plan
    3 services · web, api, worker · needs postgres, redisrunning·
  3. kapten · plan_environment(env: "staging")Plan
    plan #142 · 14 resources · est. $138/morunning·
  4. kapten · review_security(plan: 142)Plan
    0 critical · 1 medium: /metrics is publicrunning·
  6. kapten · apply_plan(plan: 142, approval: you)Act
    14/14 created · api, web, worker healthyrunning·
Animated example · sample session · tool names illustrative

Connect in one line

Claude Code
claude mcp add --transport http kapten https://mcp.kapten.io/mcp

AlsoClaude DesktopCursorVS CodeSetup guide

Protocol
MCP over HTTPS, OAuth
Clients
Claude Code, Cursor, Claude Desktop, VS Code
Permissions
Observe · Plan · Act
Crew
Eight specialist agents

Inside the product

An assistant that starts with your environment.

Interface previews
Illustrative · sample data · Early access

  • 01

    Start with a concrete question

    Give the assistant a service and an environment. Ask what is running or what changed before requesting an operation.

  • 02

    Keep the environment explicit

    Carry the project and environment into the conversation. The same service name in staging and production should not mean the same action.

  • 03

    Connect through MCP

    The intended MCP interface gives compatible assistants a structured way to work with Kapten. Client support is confirmed during early access.

  • 04

    Follow the operation to its outcome

    Explore a workflow where an assistant can carry a request through a deployment and report the resulting state.

  • 05

    Define what the agent can reach

    Agree accessible environments and actions before connecting an assistant. Keep production operations tied to a named human owner.

  • 06

    Bring your own workflow

    Start from the coding tools your team already uses. Share the client and task you want to connect when requesting access.

Ask it anything

A conversation with your environment.

Start with a concrete question: which services are running, what changed in the last deployment, or where an error appears. Give the agent an environment and a service to work with.

kapten / agentsAgent console
claude · ~/acme/payments
Welcome to Claude Codekapten mcp server · 14 tools loaded
3 tool calls · scoped to your role
Illustrative · sample data. No live cloud connection.Open the interactive demo

Early accessAgent access is an early-access capability. The conversation below is a scripted product preview; supported clients and available actions are confirmed during onboarding.

Kapten MCP · Early access

An SRE team, inside your agent.

Eight specialists, one plan to approve, nothing applied without you. The full tour (crew, tools, guardrails and architecture) lives on the MCP page.

Explore Kapten MCP
Protocol
MCP over HTTPS, OAuth
Clients
Claude Code, Claude Desktop, Cursor, VS Code
Permissions
Observe · Plan · Act
Runs in
Your cloud account and region

How the crew works

One request.
A whole crew behind it.

Your agent talks to Kapten over MCP. Kapten splits the work between specialists, runs them in parallel and brings back a single plan with the diff, the cost and the risks. Only an approved plan reaches your account.

Your agent sends a request over MCP to Kapten, who delegates to seven specialist agents. One plan returns for your approval; only approved plans act in your cloud account.Your agentClaude CodeCursor · Desktop · VS Code“Ship this repoto production.”MCPplanOrchestrationKaptenPlans and delegatesMerges the findingsAsks before actingNavigatorArchitectureShipwrightBuild & containersBosunInfrastructureHelmsmanReleasesWardenSecurity reviewLookoutObservability & incidentsPurserCostapproved plansYour cloud accountScoped role · 900 s sessionsKubernetes clusterManaged PostgresObject storageSecretsDNS & TLS

Guardrails

Observe. Plan. Then act.

Three permission tiers, granted per environment. Reading and planning never change anything; acting always goes through a plan you approved.

  • 7Observe tools
  • 6Plan tools
  • 6Act tools

  1. 01Observe

    Read the state of projects, environments, deployments, logs, metrics and costs.

    • “What is running in staging?”
    • “Why did the last deploy fail?”
    • “What did we spend last week?”

    No approval. Every call is logged.

  2. 02Plan

    Produce a dry-run: the exact resources, configuration and releases a change would create, with cost and risk.

    • “Plan a production environment for this repo”
    • “Plan a rollback of the API”
    • “Estimate a preview per pull request”

    No approval. Nothing changes until the plan is applied.

  3. 03Act

    Apply an approved plan: provision, deploy, scale, roll back, rotate a secret.

    • “Apply plan #142”
    • “Deploy api@3f2c1a to staging”
    • “Roll back web in production”

    Explicit approval per plan. Production requires a named owner.

Guide: guardrails for agents that touch production

Workflows

Start from a recipe, not a blank prompt.

Each workflow names the specialists it brings in and where it stops to ask you. Copy the prompt into your agent.

All 9 workflows

With your crew · Kapten MCP

Eight specialists, one conversation, your approval.

Kapten MCP puts a crew of specialists inside the coding agent you already use. On Agents & MCP, these are the ones on watch. They read and plan freely; anything that changes your infrastructure waits for your approval.

Meet the whole crew

  • KaptenOrchestrationLeads

    Turns your request into a plan and hands each part to the right specialist.

  • NavigatorArchitecture

    Maps the application and proposes the environment.

  • HelmsmanReleases

    Ships releases behind health checks and rolls back in one step.

  • LookoutObservability & incidents

    Investigates incidents from what changed, read-only.

  • WardenSecurity review

    Reviews exposure, secrets and access before anything ships.

  • PurserCost

    Estimates cost before you provision and finds waste after.

Your agentClaude Code · Claude Desktop · Cursor · VS Code

Try asking

What is running in staging, and which version of the API is deployed there?

  1. Kapten splits the request between Navigator, Helmsman, Lookout, Warden and Purser.
  2. Navigator maps the application and proposes the environment.
  3. Helmsman ships releases behind health checks and rolls back in one step.
  4. Lookout investigates incidents from what changed, read-only.
  5. Warden reviews exposure, secrets and access before anything ships.
  6. Purser estimates cost before you provision and finds waste after.
  7. Plan ready. Nothing changes until you approve it.

Observe & Plan · no approvalAct · your approvalIllustrative run · early access

Workflows that use it

How it works

One conversation. A clear operating context.

Each station names the crew member that owns it. Diamonds are gates: the run stops there until a person decides.

  1. Bring your client
    Kapten
    Tell us which assistant and workflow you want to connect.
  2. Agree its scope
    Warden
    Define the accessible environments and available actions.
    Gate: Scope agreed
  3. Start with a question
    Kapten
    Inspect a service or a deployment before exploring operations.
Agents & MCP, step by step. Early-access behaviour, confirmed during onboarding.
Want the long version?Connect Kapten to Claude Code, Claude Desktop and Cursor

Deep dive

The ideas behind it.

  1. Ask before acting

    Start with a concrete question: which services are running, what changed in the last deployment, or where an error appears. Give the agent an environment and a service to work with.

  2. Carry the context into delivery

    The intended workflow connects a request to its deployment and outcome, so a conversation can continue from investigation through to the release.

  3. Define the operating boundary

    Agree which environments and actions are accessible before enabling an agent. Treat production changes as operational decisions with a named owner.

Good to know

A little more
before you start.

Have a particular application or setup in mind? Bring it to the early-access conversation.

Talk through your setup
What is MCP used for here?

MCP is the interface planned for connecting compatible assistants to Kapten’s platform context and operations. It gives the assistant a structured way to work with environments and services.

Which clients can connect?

The intended clients are Claude Code, Claude Desktop, Cursor and VS Code. MCP is an open protocol, so other compliant clients may work; supported clients are confirmed during onboarding.

Can an agent change production on its own?

No. Observing and planning never change anything. Acting applies a plan you approved, and production requires a named owner. Permissions are granted per environment.

Can I connect an agent today?

Request early access and tell us which client and workflow you use. We will confirm availability and the setup for your account.

Keep exploring

One part of a bigger picture.

Build with Kapten

Start with one service.
Build from there.

Tell us what you want to deploy. We review each request and follow up about fit and onboarding on your preferred cloud.

Request early access Explore the demo

Free during early access. No card required. Cloud usage billed separately.