llms.txt
# Kodeus
> Kodeus is the operating layer for AI-native applications. You describe the agent you need, Kodeus drafts it, gives it an identity, credentials and memory, then keeps it running under the policy you set. It is currently in private preview.
Kodeus is not a prompting framework — it is the infrastructure a framework leaves you to build yourself: an isolated runtime, per-user identity and credentials, memory, policy and guardrails, observability, and usage metering. Agents can run self-hosted inside a customer's own VPC and database, or on Kodeus-managed hosting. The platform is organized into four layers: Execution (runtime, intelligence, memory), System (resources, tenancy, identity, governance), Coordination (communication, observability, scheduling, registry), and Agent Economy (usage metering, billing, settlement). Warren, a private AI fund manager, is the flagship application built on Kodeus today.
## Product
- [Homepage](https://kodeus.ai/): Overview of Kodeus, the problem it solves (credentials, guardrails, traces), how it works (describe, deploy, operate), what ships out of the box, the Warren case study, security/governance summary, and FAQ.
- [Platform](https://kodeus.ai/platform): The four operating layers (Execution, System, Coordination, Agent Economy) in detail, what's "Now" vs "Maturing" for each capability, the optional modules (Verifiable Execution, x402 payments), the build/deploy lifecycle, and how Kodeus compares to frameworks, workflow builders, model platforms, and MCP layers.
- [Developers](https://kodeus.ai/developers): CLI and SDK quickstart, the kodeus.yaml spec file format, how CLI/managed services/runtime fit together, the three credential models for attaching MCP tools (per-user env, per-user OAuth, project-shared), APIs and integrations (HTTP streaming, MCP, agent-as-MCP, agent-to-agent), and the five policy intercept points with 45 built-in guardrails.
- [Enterprise](https://kodeus.ai/enterprise): Security and governance controls for regulated deployments — data boundary, tenant isolation, identity/credential vaults, predictable usage-based cost, the governed-action lifecycle (propose → policy → approval → execute → trace), deployment models, and the optional Verifiable Execution module.
- [Use cases](https://kodeus.ai/use-cases): Categories teams build on Kodeus (autonomous markets, enterprise operations, research & analysis, customer operations, machine commerce, regulated workflows), plus a full case study of Warren, the AI fund manager built on Kodeus.
- [Pricing](https://kodeus.ai/pricing): Developer (free, self-hosted), Team ($99/mo), Business ($499/mo) and Enterprise (custom) plans, what each includes, and the usage-based billing rates for model tokens, runtime hours, tool/MCP calls, memory/storage, and marketplace revenue share.
- [Contact](https://kodeus.ai/contact): Request early access, book a demo, or talk to sales.
## Topics
- [AI agent infrastructure](https://kodeus.ai/ai-agent-infrastructure): What production AI agent infrastructure has to provide — runtime, identity, credentials, memory and policy — and how Kodeus supplies it.
- [AI agent governance](https://kodeus.ai/ai-agent-governance): Guardrails, approvals, access controls and audit evidence for running agents under enterprise policy.
- [Agent observability](https://kodeus.ai/agent-observability): Real-time monitoring, structured traces of every tool call and result, and performance analysis for agents in production.
- [LangGraph and CrewAI alternatives](https://kodeus.ai/alternatives): How Kodeus compares to composition frameworks, and why most searches for an alternative are really about the operating layer underneath.
