Skip to content
Private preview The Kodeus SDK and demo app are not public yet. Get early access
Kodeus
Platform

Everything your agent needs
to run in production

You describe the agent, Kodeus drafts it, and the same platform runs it: runtime, identity, memory, policy, observability and metering.

Agent EconomyLayer 04
Usage meteringBillingSettlement
CoordinationLayer 03
CommunicationObservabilitySchedulingRegistry
SystemLayer 02
ResourcesTenancyIdentityGovernance
ExecutionLayer 01
RuntimeIntelligenceMemory
Kodeus operating core

We use operating system as an analogy for the role Kodeus plays, not as a claim to be a kernel in the computer-science sense. Kodeus runs as software above the OS, cloud and model layers.

What Kodeus handles for you

They only work as a set. Smart but unable to act is a chatbot. Able to act but untrusted never ships. Trusted but unmeasured gets switched off when the bill arrives.

What it covers What you get
Intelligence Model abstraction, context and token optimization, and memory. Expanding to cost-, latency- and policy-based routing with provider fallback and rate control.
Execution Runtime lifecycle, per-tenant concurrency, tool orchestration, durable background routines and cancellation. Expanding to richer scheduling, event sources and horizontal scale.
Trust Authentication, encrypted secret vaults, guardrails, human approval and isolation. Expanding to role-based access and comprehensive per-action audit.
Agent economy Execution-level usage metering, the foundation for billing, payments, settlement, revenue sharing and marketplace participation.

Features, and what is available today

Pick a layer. Every capability is labelled, so you can tell what you can use this week from what is still coming.

Keeps an agent running, efficient and context-aware.

Runtime
  • Application and worker lifecycle Now
  • Per-tenant turn scheduling and concurrency limits Now
  • Fault recovery, cancellation, graceful drain Now
  • Retry and circuit-breaking at scale Maturing
Intelligence
  • Multi-model abstraction, per-agent selection Now
  • Context and token optimization, prompt cache Now
  • Cost, latency and policy-based routing Maturing
  • Provider fallback and rate limiting Maturing
Memory
  • Context-window management and compaction Now
  • Short-term and long-term memory Now
  • Tool-output persistence and retrieval Now
  • Retention and eviction Now

Isolated, secure and governable, per tenant.

Resources
  • Per-tenant memory limits, turn serialization Now
  • Usage-limit guardrails Now
  • Quotas, balancing, autoscaling Maturing
Tenancy
  • Orgs, users, apps, databases and environments isolated by tenant Now
  • Per-user credential encryption Now
Identity
  • Identity from a verified credential, never the request body Now
  • Encrypted vaults, AES-256-GCM, per-user keys Now
  • API-key issuance, rotation, revocation Now
  • Role-based access control Maturing
Governance
  • Turn-level policy at five seams Now
  • 45 built-in guardrails, enabled per agent Now
  • Human-in-the-loop approval gate Now
  • Per-action audit Maturing

Connects agents to people, events and other systems.

Communication
  • HTTP streaming API Now
  • MCP tools, agent-as-MCP Now
  • Agent-to-agent transports Now
  • Slack, Telegram, Discord, WhatsApp Maturing
Observability
  • Structured logs, OpenTelemetry traces Now
  • Per-call latency, token and cache usage Now
  • Retrievable execution histories Now
  • Replay and cost analytics Maturing
Scheduling
  • Durable wakeup scheduler that survives restarts Now
  • Bounded retry, one-shot and recurring Now
  • Cron expressions and event sources Maturing
Registry
  • Searchable catalog of tools and MCP servers Now
  • Binding and introspection Now
  • Versioning, publishing, monetization Maturing

Anchored today by usage metering: model and tool consumption captured per agent. Billing, settlement and revenue sharing build on that meter.

Usage economics
  • Model, tool and workflow consumption measured per execution Now
Creator economics
  • Revenue shared with developers, strategists and capability owners Maturing
Machine commerce
  • Paid capability discovery, micropayments, service-to-service settlement Maturing
Marketplace
  • Publish, discover, consume and monetize capabilities Maturing

Now running in the platform today. Maturing designed, shipping as the layer grows.

Optional modules

Optional add-ons

Two capabilities most teams do not need, and never pay for unless they switch them on.

Module · Verifiable Execution

Proof a counterparty can check

Anchors selected execution records (decisions, tool calls, approvals and outcomes) to a cryptographic ledger, producing proofs an auditor or regulator can verify without trusting the operator. Verification is a property applications opt into per action.

  • Cryptographic proof of execution and decision provenance
  • Tamper-evident, independently verifiable audit trails
  • Selective anchoring, per action and per policy
Module · Agent-Economy Payments

x402: payment native to how machines talk

Aligns with x402, the open standard that activates the HTTP 402 “Payment Required” status code so a service can request payment and an application can settle it, typically in stablecoins, inline over standard web protocols.

request → settle → retry
 GET /v1/dataset/tick-history
 402 Payment Required   // terms attached
 settle("0.02 USDC")      // application pays
 200 OK                  // completes inline

How an agent gets built and deployed

See it in the CLI
  1. Describe

    State the outcome in natural language, or define it through the CLI and SDK. Guided specification generation captures intent, role, goals and tools.

  2. Compose

    The platform drafts the application, assembling a model, suggesting MCP servers and scaffolding skills for you to refine.

  3. Deploy

    An isolated runtime with its own identity, credentials, memory, policies and resource limits. A managed workload, not a transient prompt sequence.

  4. Operate & govern

    The orchestrator coordinates tasks, memory, tools, background routines and human approvals, with observability and policy active throughout.

  5. Monetize

    Usage is metered at the execution level; billing, revenue share and marketplace participation build on that foundation as the agent economy layer matures.

How Kodeus compares

Keep your framework, your models and your MCP servers. Kodeus is the layer beneath them.

Approach Where the gap remains
Developer frameworks Powerful composition primitives, but teams still own runtime, security, governance, scaling, cost, distribution and monetization.
Workflow builders Useful for deterministic automation, but every flow must be explicitly designed and redeployed when conditions change.
Model platforms Provide intelligence and inference, but not the application lifecycle across tools, memory, policies, infrastructure and economics.
Tool / MCP layers Standardize connectivity, but connectivity is not orchestration, reliability or governance.
Verifiable-execution and settlement protocols Offer identity, proof or settlement primitives, but still require an application runtime and orchestration system to become useful.

Get in early

The SDK is in private preview. Tell us what you want to build and we will bring you in as access opens.