How much does Kernel cost?

Free tier (including stealth mode with residential proxies) plus usage based pay for resources pricing described as simple and transparent; specific rates not captured in retrieved materials.

What are the best alternatives to Kernel?

Based on documented capability coverage across the Agentic Index 14 point taxonomy, the closest alternatives to Kernel among browser and computer-use agents are Airtop, Tinyfish, Asteroid, Bright Data Agent Browser. Each is graded on the same public evidence, and no vendor pays for placement.

Does Kernel support MCP and tool calling?

Full under the Agentic Index verification standard. Kernel publishes public evidence of Model Context Protocol (MCP) support, tool calling, and API or SDK extensibility for connecting agents to external tools. Evidence basis: API and SDK for launching and controlling browsers, an MCP server for AI clients, framework compatibility across the automation ecosystem, and hosted agent endpoints, Kernel materials 2026-07-22. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Kernel give you tracing and observability into agent runs?

Full under the Agentic Index verification standard. Kernel publishes public evidence of tracing, observability and audit logging into agent runs, so a team can see what an agent did and why. Evidence basis: Live views, video replays, VM logs, session replays, and debugging tools let engineers see exactly what an agent saw and did in the browser, Kernel materials 2026-07-22. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Kernel have SOC 2, SSO and role based access control?

Partial under the Agentic Index verification standard. Kernel publishes some public evidence of security and identity controls, but narrower than the full set of SOC 2, single sign on (SSO) and role based access control (RBAC). Evidence basis: Isolated virtual machines per agent session and secure credential handling are documented, without third party attestations in retrieved materials, Kernel materials 2026-07-22. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Back to vendors

Kernel

Q: Who is Kernel for?

AI agent developers, startups, Series B+ enterprises building automation

Also known as: Kernel, kernel.sh, onkernel

Visit site

Browser / computer-use agentindependentVerified 2026-07-22

Entry priceFree tier + usage basedFull pricing detail

Browsers as a service for AI agents: millisecond spin ups, persistent session state, live take-control views and replays, any-framework support plus an MCP server, and an agent runtime that turns agents into HTTPS endpoints. $22M from Accel and YC.

Kernel is browsers as a service for AI agents: developers spin up cloud browsers in milliseconds (agents launch and connect in about 300ms, independently benchmarked at 4.42s full spin up versus 11.44s for Browserbase), reuse session state like cookies and authentication, and persist context across workflows so agents feel continuous rather than stateless. Browsers run in isolated virtual machines with read and write live views that let a human watch and take control mid run, plus video replays, VM logs, and debugging tools to inspect exactly what an agent saw and did.

An agent runtime hosts customer agents and turns them into HTTPS endpoints invoked on demand with a prompt. The platform supports Playwright, Puppeteer, CDP, Browser Use, Magnitude, and Stagehand, exposes an MCP server, ships stealth features (residential proxies, auto CAPTCHA solving) on the free tier, and is partnering with end websites so sites can opt in to, detect, and shape how agents interact with them. $22M from Accel, YC, and Cintrifuse; customers range from YC startups to Series B+ enterprises.

Vendor details

Canonical URL

https://www.kernel.sh

Funding status

$22M Seed + Series A led by Accel with Y Combinator and Cintrifuse Capital (October 2025); YC batch, 1,000+ customers within seven months of May 2025 launch

Company status

independent

Use cases & customers

Primary use cases

cloud browsers for computer-use agentspersistent authenticated sessions across workflowsagent hosting as HTTPS endpointsstealth web automation with proxies and CAPTCHA solvingagent run debugging via live views and replays

Target customers

AI agent developersstartupsSeries B+ enterprises building automation

Deployment options

cloud (isolated VMs per session)hosted agent runtime

Integrations

Works with any browser or web agent framework: Playwright, Puppeteer, CDP, Browser Use, Magnitude, Stagehand; APIs and SDK connect agent backends to cloud browsers for navigation, clicks, and form filling; MCP server for AI clients; residential proxies and auto CAPTCHA solvers.

Sources & related URLs

Research sources

https://www.kernel.sh https://siliconangle.com/2025/10/09/kernel-raises-22m-power-browser-infrastructure-ai-agents/ https://ycombinator.com/launches/O5f-kernel-crazy-fast-browser-infrastructure https://techstackups.com/comparisons/browserbase-vs-kernel-cloud-browser-automation/

Agentic Index coverage score

8.0 / 14 capabilities · 57%

Integrations & Tool Calling	Full
Works with any browser or web agent framework (Playwright, Puppeteer, CDP, Browser Use, Magnitude, Stagehand) with APIs connecting agent backends to browsers for navigation, clicks, and form filling, Kernel materials 2026-07-22
Workflow Orchestration	Partial
The agent runtime hosts and executes customer agents as invocable endpoints and persists context across workflows, though orchestration logic itself lives in the customer's agent, Kernel materials 2026-07-22
Knowledge Grounding & RAG	Unable to verify
No knowledge grounding or retrieval layer documented; Kernel is execution infrastructure beneath the agent's intelligence, Kernel materials 2026-07-22
Human Oversight & Guardrails	Partial
Human in the loop interactions are supported by design, with read and write live views letting a human take control of a running browser when needed, Kernel materials 2026-07-22
Security, Identity & Governance	Partial
Isolated virtual machines per agent session and secure credential handling are documented, without third party attestations in retrieved materials, Kernel materials 2026-07-22
Observability & Auditability	Full
Live views, video replays, VM logs, session replays, and debugging tools let engineers see exactly what an agent saw and did in the browser, Kernel materials 2026-07-22
Memory & State Persistence	Full
Persistence is a core pitch: agents reuse session state like cookies and authentication and persist context across workflows so they feel continuous rather than stateless, Kernel materials 2026-07-22
Deployment & Data Residency	Partial
Cloud service with per session VM isolation; self hosting and data residency options not documented, Kernel materials 2026-07-22
Prebuilt Agents, Templates & Packs	Unable to verify
No prebuilt agent or template library documented; customers bring their own agents, Kernel materials 2026-07-22
Triggers & Channel Coverage	Partial
Hosted agents become HTTPS endpoints invoked on demand with a prompt, an API trigger surface without scheduled or event machinery documented, Kernel materials 2026-07-22
Model Flexibility & Routing	Unable to verify
Model agnostic in practice since customers bring their own agents, but no model access or routing features are documented as platform capabilities, Kernel materials 2026-07-22
APIs, SDKs & MCP Extensibility	Full
API and SDK for launching and controlling browsers, an MCP server for AI clients, framework compatibility across the automation ecosystem, and hosted agent endpoints, Kernel materials 2026-07-22
Testing, Debugging & Optimization	Partial
Session replays and debugging tools support inspecting and diagnosing agent behavior, short of structured testing or evaluation frameworks, Kernel materials 2026-07-22
Browser & Computer Use	Full
The product exists to give computer use agents cloud browsers, with headful mode, stealth residential proxies, auto CAPTCHA solving, and work on enabling computer use agents industry wide, Kernel materials 2026-07-22

The Agentic Index coverage score grades every vendor Full, Partial or Unable to verify against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Free tier + usage based

usage based (browser resources)

Free tierTrial available

What is public

The model is public (free tier, then transparent usage based pay for resources); the specific per resource rates were not captured in retrieved materials.

Variable cost rationale

Usage based resource billing scales with browser hours and sessions; the vendor emphasizes predictability but costs track agent volume.

Sales call required

No, self serve available

Free / trial

Free tier (includes stealth mode)

Agentic Index verified 2026-07-22

Related vendors

Alternatives to Kernel

The closest documented capability profiles to Kernel among browser and computer-use agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

Airtop10.0 / 14Adds documented Prebuilt Agents, Templates & PacksKernel vs Airtop →
Tinyfish8.0 / 14Adds documented Knowledge Grounding & RAG
Asteroid9.5 / 14Fuller documented coverage on Workflow Orchestration and Human Oversight & Guardrails
Bright Data Agent Browser6.5 / 14Adds documented Prebuilt Agents, Templates & Packs
Visualping8.5 / 14Adds documented Knowledge Grounding & RAG and Prebuilt Agents, Templates & Packs
Autotab7.0 / 14Adds documented Knowledge Grounding & RAG

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Compare Kernel with other vendors →Compare all browser and computer-use agents →See where Kernel ranks →

Data confidence: high

Head to head

Kernel vs Airtop Kernel vs Browserbase Kernel vs Anchor Browser