When should you choose Kubiya?

Your platform team will build its own agents and teams under OPA policies. Policies should be enforced as code, with RBAC, ABAC and SSO. Memory should persist across agent runs. Graded by Agentic Index against the same capability axes from each vendor's own published materials. No vendor pays for placement.

When should you choose RunWhen?

You want about 900 troubleshooting Skills from a public registry, ready to run. Assistants should work in the background on events and schedules as well as in chat. Outcome based contracts suit how you buy. Graded by Agentic Index against the same capability axes from each vendor's own published materials. No vendor pays for placement.

How much do Kubiya and RunWhen cost?

Kubiya: Startup from $15,000 a year ($1,250 a month). SMB from $72,000 a year. Enterprise on request.. RunWhen: Self hosted license from $25,000 a year for 5,000 tickets. A PoC is $30,000 one time, and outcome based contracts start at $150,000.. Graded by Agentic Index against the same capability axes from each vendor's own published materials. No vendor pays for placement.

Agentic Index

Kubiya vs RunWhen (2026)

Both vendors last verified September 29, 2026

Kubiya and RunWhen both orchestrate agents for operations work, and both can be self hosted. That verdict is the Agentic Index coverage score, graded from each vendor's own published materials.

Kubiya is an agentic engineering platform for DevOps and platform teams, running agents and teams on a self hostable control plane with a context graph, cognitive memory, OPA policies, scheduled and webhook jobs, and an API, SDK and MCP server. Startup is from $15,000 a year. RunWhen's assistants investigate issues and run troubleshooting Skills from a public registry of about 900, in chat or in the background on events and schedules, hosted, hybrid or self hosted. Self hosting is licensed from $25,000 a year, covering 5,000 tickets. Kubiya adds RBAC, ABAC, SSO and mTLS, and records each task's agent, runtime and every tool call. Choose Kubiya for a platform to build governed agent teams. Choose RunWhen for a large library of ready troubleshooting Skills.

On the Agentic Index production ops ranking, neither Kubiya nor RunWhen clears the bar, which asks for all five resolution loop capabilities documented in full. Kubiya does not document human oversight and guardrails in full; RunWhen does not document observability and auditability in full, nor human oversight and guardrails. 15 of the 38 vendors in the lane clear it. See the production ops ranking

This comparison is published by Agentic Index, an independent agentic AI vendor research platform. Kubiya and RunWhen are each graded against the same 14 capability Agentic Index taxonomy, from the vendor's own public materials under the Agentic Index verification standard, alongside 1,084 researched vendors. No vendor pays for placement and no vendor has reviewed this page. How this evidence is graded

At a glance

Kubiya

Category: SRE / DevOps agent
Entry price: Startup from $15,000 a year ($1,250 a month). SMB from $72,000 a year. Enterprise on request.
Free or trial: None published
Pricing confidence: public partial

RunWhen

Category: SRE / DevOps agent
Entry price: Self hosted license from $25,000 a year for 5,000 tickets. A PoC is $30,000 one time, and outcome based contracts start at $150,000.
Free or trial: None published. Open source Skills are published in a public registry
Pricing confidence: public exact

Keep exploring

Choose Kubiya if

Your platform team will build its own agents and teams under OPA policies.
Policies should be enforced as code, with RBAC, ABAC and SSO.
Memory should persist across agent runs.

Choose RunWhen if

You want about 900 troubleshooting Skills from a public registry, ready to run.
Assistants should work in the background on events and schedules as well as in chat.
Outcome based contracts suit how you buy.

Feature	K Kubiya	R RunWhen
Action & orchestration
Integrations & Tool Calling Ability to connect agents to real systems through native integrations, OAuth-authenticated actions, custom tools, APIs, webhooks, or MCP-compatible tools.
KubiyaIntegrations & Tool Calling Connectors let agents act on outside systems by creating resources, modifying configurations, sending messages and executing workflows, across AWS, GitHub, Jira, Slack, Kubernetes and OAuth integrations to 40+ services, with custom integrations in the Control Plane API. Sourcedocs.kubiya.ai/infrastructure/integrationsread 2026-09-29
RunWhenIntegrations & Tool Calling Assistants investigate issues, run tasks and recommend actions in the customer's environment through Skills drawn from 896 Skill tools in the public registry for infrastructure, OSS, logs and metrics. A Read and Write permission level is available, and the license covers all ticket system integrations plus Slack and Teams. Sourcedocs.runwhen.com/docs/use/engineering-assistantsread 2026-09-29
Workflow Orchestration Ability to sequence, branch, retry, route, and combine deterministic workflow nodes with autonomous agent steps.
KubiyaWorkflow Orchestration Teams are a coordination layer over multiple agents that share context, tools, credentials and guardrails, and a task aimed at a team coordinates its member agents through the workflow. The Control Plane API adds plan generation, plan execution and workflow endpoints. Sourcedocs.kubiya.ai/core-concepts/teamsread 2026-09-29
RunWhenWorkflow Orchestration Several assistants, each with its own persona, permission level and confidence thresholds, choose which Skills to suggest or run. Foreground agents answer chat, and background agents run on events and cron. Customers author and commit SLXs through the Tool Builder, and these run on a schedule or on demand. Sourcedocs.runwhen.com/docs/use/engineering-assistantsread 2026-09-29
Triggers & Channel Coverage How agents wake up and where they work: schedules, webhooks, message events, CRM events, inbox events, chat, email, voice, and collaboration tools.
KubiyaTriggers & Channel Coverage Background Jobs run automatically on a set schedule, from presets or a custom cron expression, or when an external system such as Jenkins, GitHub or an internal app calls their webhook. Sourcedocs.kubiya.ai/core-concepts/background-jobsread 2026-09-29
RunWhenTriggers & Channel Coverage Background Agents run in response to events and in cron jobs, committed SLXs run on a schedule or on demand, and the platform is licensed by the tickets it works, beside chat in Slack and Teams. Sourcerunwhen.comread 2026-09-29
Knowledge & context
Knowledge Grounding & RAG Ability to ground agent behavior in company data through document ingestion, retrieval, external knowledge APIs, semantic search, or RAG layers.
KubiyaKnowledge Grounding & RAG The Context Graph ingests the customer's AWS, GitHub, GitLab, Azure, GCP, Kubernetes and custom sources on a continuous, hourly, daily or cron sync, persists the entities and relationships, and exposes graph, semantic and intelligent search to agents. Sourcedocs.kubiya.ai/web-interface/context-graph/data-ingestionread 2026-09-29
RunWhenKnowledge Grounding & RAG Agents reference a database of prior runs, and workspace intelligence lets them search issues and browse the SLXs generated for the customer's discovered resources. That is a persisted store of findings about the customer's estate that agents can query. Sourcedocs.runwhen.com/docs/use/mcp-serverread 2026-09-29
Memory & State Persistence Ability to persist context across a run, conversation, workflow, user, team, or longer-term memory layer.
KubiyaMemory & State Persistence Cognitive Memory is a store that agents write to with store_memory and users add to through Compose, scoped per user, organization or role (organization by default). Kubiya states no lifetime, retention or delete path for it. Sourcedocs.kubiya.ai/core-concepts/cognitive-memory/overviewread 2026-09-29
RunWhenMemory & State Persistence There is no memory with a set scope or lifetime. The database of prior runs that agents reference stores run results, the application's own data, not a memory layer. Sourcedocs.runwhen.com/docs/use/engineering-assistantsread 2026-09-29
Control & trust
Human Oversight & Guardrails Approval steps, consent checkpoints, escalation rules, structured guardrails, policy constraints, and pause/resume controls.
KubiyaHuman Oversight & Guardrails OPA policies attached to environments, teams or agents constrain what agents may run, such as business hours windows and required labels, and a starter policy, described in a single line, asks for explicit confirmation or security signals before sensitive actions. A task can also wait in a Requires Input state. Apart from request, grant and deny access endpoints in the API reference, there is no approval step before an action commits. Sourcedocs.kubiya.ai/core-concepts/policiesread 2026-09-29
RunWhenHuman Oversight & Guardrails Each assistant carries a Read Only or Read and Write permission level, access limited to chosen users, and a Run Confidence threshold that decides what it runs and what it only suggests. Changes follow a Git based workflow. These are constraints the customer sets. There is no approval step before a task runs, and RunWhen says it works with a human in the loop but names no mechanism. Sourcedocs.runwhen.com/docs/use/engineering-assistantsread 2026-09-29
Security, Identity & Governance RBAC, SSO, auditability, encryption, least-privilege tool access, compliance posture, and data handling policy.
KubiyaSecurity, Identity & Governance Kubiya offers RBAC, ABAC, SSO, social login and an audit trail, along with policy as code, mTLS and OAuth2. A SOC2 mark appears in the Composer page's security section, but no report type, report or trust page is published. Sourcekubiya.ai/pricingread 2026-09-29
RunWhenSecurity, Identity & Governance Access runs on role based access control, authentication and authorization, audit logging and per assistant permission levels. The SOC 2 Type II audit is still ahead. RunWhen will undergo it every year once the product reaches general availability, so no attestation is held yet. Sourcerunwhen.com/security-informationread 2026-09-29
Observability & Auditability Traces, logs, execution histories, metrics, audit events, and debugging detail for production agent behavior.
KubiyaObservability & Auditability Each task records its agent, runtime and execution ID, along with live agent reasoning and actions, each tool invocation with its status, and full logs and artifacts. That history is kept for audit, and the API exposes traces and spans per execution. Sourcedocs.kubiya.ai/core-concepts/task-kanbanread 2026-09-29
RunWhenObservability & Auditability Run sessions can be reviewed and the platform has audit logging, but RunWhen does not say what a run session records of the Skills an assistant ran, their commands or outputs. Sourcedocs.runwhen.com/docs/use/mcp-serverread 2026-09-29
Deployment & Data Residency Deployment modes and options, including SaaS, dedicated cloud, VPC, on-prem, hybrid, local runtime, and self-hosting.
KubiyaDeployment & Data Residency The Control Plane can be self-hosted on the customer's own infrastructure, on local machines, cloud VMs (AWS, GCP, Azure), Kubernetes clusters or bare metal, with workers executing in the customer's environment. Running AI Teammates in the customer's cluster and local runners are listed plan features. Sourcedocs.kubiya.ai/cli/control-plane-self-hostingread 2026-09-29
RunWhenDeployment & Data Residency There are three options. RunWhen Hosted runs in Google Cloud Canada and United States regions. Hybrid runs scripts in the customer's clusters through the Local agent. Self Hosted puts the full stack in the customer's VPC and suits air gapped and very high compliance environments. Sourcerunwhen.com/security-and-deploymentread 2026-09-29
Solution readiness
Prebuilt Agents, Templates & Packs Ready-made workflows, packaged employees, templates, blueprints, industry solutions, and role-specific agents that reduce time-to-value.
KubiyaPrebuilt Agents, Templates & Packs Default agents ship in pre-built Terraform modules for a customer to adopt and customize, a Deployer for production deployment and a Monitor for monitoring and alerting, plus a default DevOps team. Sourcedocs.kubiya.ai/terraform/modulesread 2026-09-29
RunWhenPrebuilt Agents, Templates & Packs A public Skills Registry holds 896 Skill tools for infrastructure, OSS, logs and metrics. Each is a ready made troubleshooting task an assistant can pick up, reviewed before agents use it, and runnable as a standalone CLI tool. Sourcerunwhen.comread 2026-09-29
Platform extensibility
Model Flexibility & Routing Ability to work across multiple foundation models, route tasks to different models, or let buyers bring their own providers and keys.
KubiyaModel Flexibility & Routing Workers can send model calls through the customer's own providers instead of the default Control Plane gateway, over LiteLLM's 100+ providers (AWS Bedrock, Azure OpenAI, Vertex AI, OpenAI, Anthropic, Mistral, or local Ollama and vLLM), and Bring Your Own Model is a listed plan feature. Sourcedocs.kubiya.ai/cli/custom-llm-gatewayread 2026-09-29
RunWhenModel Flexibility & Routing The platform supports bring your own LLM endpoint, so the customer chooses the model the assistants use. Sourcerunwhen.comread 2026-09-29
APIs, SDKs & MCP Extensibility Composability layer: stable APIs, SDKs, MCP tool consumption/serving, custom tools, and integration into internal systems.
KubiyaAPIs, SDKs & MCP Extensibility The Control Plane API (REST, with an OpenAPI spec) covers agents, teams, executions, jobs, policies, models and the context graph, beside a Python SDK, a CLI and a Terraform provider. The Kubiya MCP server (@kubiya/control-plane-mcp-server, API key auth) exposes create, read, update and delete tools over agents, teams, jobs, policies and workflows. Sourcedocs.kubiya.ai/mcp/overviewread 2026-09-29
RunWhenAPIs, SDKs & MCP Extensibility RunWhen's own MCP server, still in beta (hosted at mcp.beta.runwhen.com/mcp or local over stdio, bearer token or OAuth), exposes workspace search, chat rules and Tool Builder tools that commit SLXs to run on a schedule, and there is an API reference. Sourcedocs.runwhen.com/docs/use/mcp-serverread 2026-09-29
Testing, Debugging & Optimization Testing, debugging, scoring, retries, fallbacks, quality gates, and optimization loops for improving agent workflows before and after deployment.
KubiyaTesting, Debugging & Optimization Policies are validated through the CLI before deployment, and the API validates policy, runtime and skill configuration, which confirms that a change is well formed but does not test how the agent behaves. Analytics report usage and cost, and there is no evaluation harness, scored tests or improvement loop over agent behavior. Sourcedocs.kubiya.ai/core-concepts/policies-use-casesread 2026-09-29
RunWhenTesting, Debugging & Optimization Skills built in the Skill Building Harness go through multiple levels of AI code review before an agent can pick them up, and they are tested against the RunWhen contract and live infrastructure, which validates the tools. User feedback that shows where new tools should be prioritized is mentioned only in passing. There is no evaluation of the assistants' own investigations. Sourcerunwhen.comread 2026-09-29
Specialist automation
Browser & Computer Use Browser, desktop, or remote/local computer control for workflows that cannot be handled through stable APIs alone.
KubiyaBrowser & Computer Use Agents act through connectors, skills (shell, Docker, Python, file system) and sandboxed execution, and do not control a browser, desktop or computer. Sourcedocs.kubiya.ai/infrastructure/integrationsread 2026-09-29
RunWhenBrowser & Computer Use Assistants run CLI and API Skills in the customer's environment. There is no browser, desktop or computer control. Sourcedocs.runwhen.com/docs/use/engineering-assistantsread 2026-09-29

Pricing snapshot

Sourced from the Index pricing dataset · open each vendor's profile for full detail.

Pricing	K Kubiya	R RunWhen
Entry price Lowest public entry point	Startup from $15,000 a year ($1,250 a month). SMB from $72,000 a year. Enterprise on request.	Self hosted license from $25,000 a year for 5,000 tickets. A PoC is $30,000 one time, and outcome based contracts start at $150,000.
Pricing confidence How public the numbers are	Public, partial	Public, exact
Billing Primary billing axis	Annual plan by tier, sized by AI Teammate profiles, advanced and fast model token allowances, concurrent tasks, integrated accounts, local runners, scheduled tasks and webhooks.	An annual self hosted license priced by ticket and chat volume, a one time PoC, and outcome based contracts for unlimited sessions.
Variable cost Workload / overage exposure	Medium variable cost	Medium variable cost
Free tier / trial Try before you buy	No free tier	No free tier
Buying motion Self-serve vs sales call	Mixed	Sales call
Pricing verified When this pricing was last checked	Sep 29, 2026	Sep 29, 2026

Product updates

Newest first, each linked to its source.

RunWhen

September 28, 2026 · MCP / tool calling / API

RunWhen published the first tagged releases of runwhen-capability, the open source Python SDK its agent skills are written against, along with rwtask, a task host that runs each skill as a warm executor and doubles as a local command line tool.
Source

All RunWhen updates →

RunWhen's proof of concept is $30,000 one time, and its outcome based contracts start at $150,000.

Kubiya profile RunWhen profile Compare the full category See SRE and DevOps agents ranked Build a custom comparison