BUILD

Build Scalable,
Production-Ready Agents

Q: Do we have to switch frameworks, or will we be locked in?

No, and no. The Komodor Agentic Operations Platform is framework-agnostic. Keep building with the harness and coding agents you already use, and bring the results to the platform as they are. There is no new DSL to learn. Komodor adds the shared infrastructure around the agent: identity, governance, memory, evaluation and orchestration. It doesn’t replace how you write the agent. Docs link: How building works

Q: Do we need to write code to build an agent?

Not necessarily. You can deploy a ready-made agent from the catalog, or create one in the console from a template. You can also ask a coding agent such as Claude Code, Codex CLI or Cursor to build it for you over Komodor’s MCP server. When you want full control, write it against the SDK. Every path produces the same governed agent, and each one starts as a draft before it goes live. Docs link: How building works

Q: What agents come out of the box?

Komodor ships 50+ specialist agents, already used in enterprise production environments. They cover Kubernetes, cloud providers (AWS, Azure and Google Cloud), APM tools such as Datadog and Grafana, code review, remediation and orchestration. You deploy them from the agent catalog and configure their credentials, triggers and permissions. You can run them alone or alongside your own agents in the same workflows. Docs link: Agent catalog

Q: How do we add our own runbooks, procedures and team knowledge?

There are two ways. Package runbooks and procedures as skills. Skills are versioned, reusable units of know-how that you attach to any agent, including Komodor’s out-of-the-box agents, and updates reach agents without a restart. Upload runbooks, postmortems, service docs and policies to the knowledge base. Agents search the knowledge base while they work and cite what they used. Docs link: Skills

Q: Can agents connect to our in-house tools and internal APIs?

Yes. Komodor includes built-in integrations for common tools such as AWS, Azure, Google Cloud, Datadog, Dynatrace, New Relic, Grafana, GitHub, GitLab, PagerDuty and Jira. For in-house systems, register your own MCP servers or OpenAPI documents with the MCP Gateway, and their operations become tools your agents can call. You decide which agents can reach each tool, and credentials stay out of agent code. Docs link: MCP Gateway

Q: Can we use Komodor's agents and our own in the same workflow?

Yes. Any step in a workflow can be filled by an agent from Komodor’s catalog or by one you built or imported. Both are first-class agents on the platform, so they share the same context and hand results to each other. They run under the same governance and evaluation. Many teams start with Komodor’s prebuilt workflows and swap in their own agents where they already have something that works. Docs link: Modules & workflows

Q: What's the difference between a skill, an agent, a workflow and a module?

A skill is a reusable piece of know-how, such as a runbook or procedure, attached to an agent. An agent is the governed identity that does the work: its instructions, tools, permissions and history. A workflow is a sequence of steps, each handing a task to one or more agents. A module packages workflows around one operational outcome, such as incident management or Kubernetes cost. Docs link: Concepts & glossary

Build production-ready agents your way with Komodor. Connect to the harness of your choice with our SDK and bring what you’ve already built – or start from 50+ curated, proven agents available out of the box. Every agent you create or import gets the same guardrails, budget controls and observability – wrapped and ready to go in minutes.

Book Demo

How do I get from a working agent to one that’s ready for production?

An agent on your laptop and an agent that’s reliable and effective in production aren’t the same thing. Getting there means deciding on the right model, the harness, the MCP servers, the permissions, the owner, all wired by hand. What works as a proof-of-concept often fails against real data in a dynamic environment. Komodor gets you into production in minutes. Install the SDK on the harness you already use, and it wires in automatically, with no vendor lock-in.

Create Your Own  Powerful Agents

Whether you start with one of our ready-to-go domain specialist agents, or build your own using the provided blueprints and SDK, Komodor takes the time and effort out of building production-ready agents.

Explore docs

Get started immediately with our curated catalog of 50+ production-proven, optimized agents built to handle common operational tasks.
Build with your coding agent of choice, use out-of-the-box agent blueprints, or write directly against the Komodor SDK.
New agents automatically inherit your team’s tool and MCP connections, credentials, and prompt templates, with no need for additional setup or authentication.
Deploy self-hosted via Helm or straight to Komodor Cloud – same agent, same behavior.
Every agent — pre-built or your own — inherits the same guardrails, budgets, observability and optimization as Komodor’s own, from the first run.

Import Agents You’ve Already Built

Already building agents to help carry out operational tasks? Whatever framework they’re built on, bringing them to Komodor makes every agent you already have visible, governable and measurable — same code, same harness.

Bring existing agents in minutes. Import agents built on the coding agent of your choice, including Claude Agent SDK, LangChain, Google ADK, or AWS AgentCore and keep the logic you’ve already written.
Every imported agent is a first-class citizen, picking up the same governance, orchestration, and evaluation as an agent built from scratch on Komodor.

Get Started Fast with
the Agent Catalog

Choose from 50+ out-of-the-box specialist agents already validated, optimized and battle-tested at scale in enterprise production environments.

Explore Agents Catalog

Ready-to-go agents for common operational domains and  tasks: Observability, Cloud Agents, Kubernetes, Cost Optimization agents, DB Agent, Code agents.
Continuously evaluated and optimized across 100,000 production workloads.
Use your preexisting custom agents and Komdor agents in the same workflows.

Inherit the Same Rails,  From Day One

However you build or import it, every agent on Komodor picks up the same infrastructure automatically — no separate setup per agent.

Governance

Permissions, guardrails, brokered credentials and budget controls, enforced as the agent runs, plus approval gates for any higher-risk actions.
Integrations

Set up MCP tool servers and connectors once, shared automatically by every agent after the first.
Memory & Context

All agents share the same memory and knowledge graph — improving speed and performance on recurring issues over time.
Evaluation

Every agent run is graded under the same evaluation framework from its first run.
Optimization

Agents get cheaper and more accurate over time, with model, prompt, and tool choices continuously tuned against the customer’s own evaluation criteria.

The Gap Between an Agent and a Reliable Operational Process

An incident response workflow orchestrates agents across multiple steps from trigger to outcome.

With Komodor, agents get the step-by-step orchestration needed to solve real operational pain for Ops teams. Three solutions — AI SRE, Cost Optimization, and AI Software Operations — each package their own solution modules, containing the agents, steps, tools and integrations required, already validated and ready to run.

Ingest & Produce

Synthesize & Analyze

Signal Agent

Investigate

Agent A VS Agent B

Remediate

Proposer Executor

Verify

No agent needed

Notify

Out-of-the-box agent
Bring-your-own agent
VS Competing agents
Tandem agents

Build

Build or import your own custom agents using the Komodor SDK, or choose from more than 50 ready-to-go specialist Ops agents available.

Run

Wherever it runs, every agent shares the same fleet visibility, memory, and knowledge graph the moment it’s live.

Optimize

However it was built, every agent is scored under the same evaluation engine from its first run.

Govern

Whatever it’s allowed to touch, every agent inherits the same RBAC and credential-brokered access as your human users.

Can we bring agents we've already built without rewriting them?

Yes. Komodor imports an existing agent by generating a thin wrapper that registers it with the platform. Your agent’s logic stays in your code and keeps running where it runs today. Ready-made adapters cover Claude Agent SDK, LangChain and Google ADK, and any other framework can connect through the worker API. From its first run, the imported agent is visible, governed and measured alongside Komodor’s own agents.

Docs link: Import an existing agent

Do we have to switch frameworks, or will we be locked in?

Do we need to write code to build an agent?

What agents come out of the box?

How do we add our own runbooks, procedures and team knowledge?

Can agents connect to our in-house tools and internal APIs?

Can we use Komodor's agents and our own in the same workflow?

What's the difference between a skill, an agent, a workflow and a module?

Ready to build agents that deliver
real operational outcomes?

Book Demo

Build Scalable,
Production-Ready Agents