NEWKennen v2.0 — long-running agents, now in GA

Ship AI agents with Kennen.

Kennen is the runtime for production AI. Sandboxed execution, sub-second routing across 14 providers, continuous evals — without operating a single GPU.

Free tier · No card required
kennen — zsh

Trusted by 18,000+ teams shipping AI in production

Why Kennen

Six things you'd build yourself.
Now you don't have to.

The boring infra layer between your code and the model. Battle-tested by teams running 4B+ tokens daily.

Sub-second inference

Kennen routes across 14 model providers and picks the fastest one for your task. p50 of 380ms on 200k-token contexts.

Sandboxed execution

Every agent runs inside a hermetic Firecracker microVM. Network rules, secrets, and disk snapshots managed declaratively.

Composable agents

Build agents from primitives: tools, memories, evals. Chain them with our DAG runtime — no orchestration glue code.

Continuous evals

Every PR runs your eval suite against 4 frontier models. Regress on accuracy, latency, or cost — fail the build.

Edge-native

Workers run in 32 regions. Your agents wake within 8ms of the user — no cold starts, no warm pools.

BYO keys & data

Bring your own provider keys, your own VPC, your own retention policy. SOC 2 Type II, HIPAA, EU residency.

Composable runtime

Wire agents with a visual graph.

Or in code. Or both. The same DAG that runs in production renders live in your dashboard, with token streams flowing in real time.

webhookgithub.pr
fetchdiff.txt
classifysonnet-4
reviewopus-4
linthaiku
aggregateevals
commentgithub.api
0.0B
Tokens processed daily
0ms
p50 inference latency
0.00%
Uptime SLA
0+
Agents in production
Loved by builders

The teams shipping fastest
are already on Kennen.

"We replaced our entire orchestration layer with Kennen in a weekend. 60% lower latency, 40% lower cost, half the code."

MO
Maya Okafor
Staff Eng · Halcyon

"Kennen is the first agent platform that actually feels like infra. Deterministic, observable, boring in the best way."

DM
Diego Marín
CTO · Cobalt

"Our eval-on-PR setup catches regressions we used to find in prod. Kennen paid for itself in the first month."

PR
Priya Raman
Head of AI · Linework

"We replaced our entire orchestration layer with Kennen in a weekend. 60% lower latency, 40% lower cost, half the code."

MO
Maya Okafor
Staff Eng · Halcyon

"Kennen is the first agent platform that actually feels like infra. Deterministic, observable, boring in the best way."

DM
Diego Marín
CTO · Cobalt

"Our eval-on-PR setup catches regressions we used to find in prod. Kennen paid for itself in the first month."

PR
Priya Raman
Head of AI · Linework

"We replaced our entire orchestration layer with Kennen in a weekend. 60% lower latency, 40% lower cost, half the code."

MO
Maya Okafor
Staff Eng · Halcyon

"Kennen is the first agent platform that actually feels like infra. Deterministic, observable, boring in the best way."

DM
Diego Marín
CTO · Cobalt

"Our eval-on-PR setup catches regressions we used to find in prod. Kennen paid for itself in the first month."

PR
Priya Raman
Head of AI · Linework
Pricing

Start free.
Pay for what you ship.

Hacker
$0/mo

For solo devs and side projects.

  • 1M tokens/month
  • 5 agents
  • Community Discord
  • Public eval reports
Most popular
Team
$99/mo

For shipping teams.

  • 100M tokens/month
  • Unlimited agents
  • Priority routing
  • GitHub Actions integration
  • Email support
Scale
$499/mo

For platform teams.

  • Custom token limits
  • Dedicated capacity
  • BYO keys & VPC
  • SSO + audit logs
  • Solutions architect
FAQ

Everything else.

Anthropic, OpenAI, Google, Meta, Mistral, Cohere, DeepSeek, Qwen, plus your own self-hosted endpoints. Routing is automatic; you can also pin per-agent.

Ship your first agent in 60 seconds.

Free tier includes 1M tokens. No credit card. Deploy from your terminal in one command.