AI Agent SDK & Developer Platform | Build AI Agents – Kodeus
Developers

Start in an afternoon

Create the agent in the Console or author it as a spec with the open source SDK. Run it locally, then deploy the same thing to your cloud or ours. The docs have the rest.

Request early access →Book a demoDocs
KodeusCONSOLE
+ New agent
Agents
Tools and credentials
Knowledge
Skills
Schedules
Templates
ACCOUNTAgents3 agents on this org.
RefreshNew agent
Ssupport-agentrunning
anthropic/claude-sonnet-4Resolves support requests in Zendesk and the billing system. Refunds over the cap wait for a person.
your cloud3 users
Rria-adviserrunning
generic.deepseek/deepseek-v4Meeting packs, drift flags and client emails for a named adviser. Rebalances wait for a yes.
kodeus cloud3 users
Lledger-reconrunning
openai/gpt-5Nightly reconciliation across ledgers. Mismatches above threshold go to a person.
local3 users
BUILDER“Triage our support tickets and draft the easy replies”thought for 2s · attached Zendesk · asks how far it should go on its own
The shape of it

You write the spec. The runtime runs it.

A Kodeus agent is a kodeus.yaml file plus the skills you author. The SDK turns it into artifacts in your own database, and the runtime runs it on your infrastructure. You own the data and the infrastructure.

Your machine

Kodeus SDK

Validates kodeus.yaml, scaffolds skills and prompts, launches the runtime locally.

spec →← artifacts
Kodeus Cloud

Synthesis service

Turns the spec into prompts and artifacts. Operated by Kodeus. Never in the serve path.

artifacts written ↓
you own
Your database

Specs, artifacts, memory, traces

Postgres you already run. Everything the runtime needs to serve lives here.

reads
you own
Your infrastructure

Kodeus runtime

Runs the agent per user with identity, policy, memory and traces. Laptop, VPC, or airgapped.

kodeus.yaml

One file you can read

Model, tools, skills, memory, policy and limits, in one file. Policy is declared, not buried in a prompt.

{{ l.text }}
{{ c.key }}{{ c.text }}
Deploy

The same runtime on your laptop and in production

Run it locally with the same policy engine as production. When you deploy from the Console, nothing about the agent changes. Only where it runs.

LocalYour cloudKodeus CloudAirgapped
Where it runs, on the Platform page →
Evals

A correct run is a test

Record a correct run. Replay it after a prompt edit or a model bump. Same path, ship it. Different path, you found the drift before your users did.

recorded · claude-sonnet-4live
1 crm.lookup
2 policy · pass
3 refunds.create · $120
4 email.send
recorded · 1.6s · $0.0031
replay · gpt-5live
1 crm.lookup
2 policy · pass
3 hold · asks approval
4 email.send · not reached
diff at step 3 · review before ship
FAQ

Questions developers ask

{{ f.a }}

Private preview

Stop building the platform. Start shipping the product.

Tell us what you are building and we will get you into the preview as soon as we can support your use case properly.

Request early access →Book a call