Every capability you need.
None you don't.

OLAV is not a chatbot wrapper. It is a governed, audited, self-improving operations platform purpose-built for teams who need reliability, traceability, and control.

Core Capabilities

Any API → Agent Tool

Point OLAV at any OpenAPI schema URL and every endpoint becomes a natural-language-accessible tool — instantly, with zero code.

🏗️

Creator Agent

6-step guided workflow. Upload a schema, define a Skill, test, validate, and deploy — in under 2 minutes. No framework knowledge required.

🧠

Self-Improving Loop

Failure patterns from every run are automatically captured and converted into execution constraints. Agents get measurably better over time.

🚀

2000x+ Cache Speedup

Semantic deduplication + exact-match cache layers. Repeated queries return in milliseconds. LLM tokens are spent only when the request is genuinely novel.

🔌

HTTP API & CLI

Full REST API surface alongside the OLAV CLI. Integrate into existing toolchains, CI pipelines, or dashboards without rewriting workflows.

📦

Skill Registry

Discover, share, and version Skills across your team. One registry entry serves all users on the platform — no per-user setup required.

Security Architecture

Every agent execution path is gated by four mandatory control layers. These are architectural constraints — not configuration options. No setting can bypass the DuckDB read-only sandbox or credential redaction.

Layer 0 AAA
  • Token, LDAP, and OIDC authentication
  • Role-based access control (RBAC)
  • Full audit trail to DuckDB
Layer 1 Middleware
  • HITL interception for dangerous commands
  • Semantic memory injection per session
  • Configurable approval gates
Layer 2 Sandbox
  • Pre-execution command scan
  • DuckDB read-only enforcement (cannot be disabled)
  • Network namespace isolation per agent
Layer 3 Output
  • Automatic credential and secret redaction
  • SSE + JSON structured output
  • HttpOnly cookie session binding

Multi-User Operations

👥

Multi-User & Concurrent Sessions

Multiple operators share one platform instance with fully isolated sessions. No request bleed-through between users.

📋

Complete Audit Trail

Every CLI command, agent decision, and tool call is persisted in DuckDB with timestamps, user identity, and results. Exportable for compliance.

🔐

LDAP / OIDC Integration

Plug into your existing directory. Map LDAP groups or OIDC claims to OLAV roles. No shadow user database to maintain.

🏷️

Workspace Isolation

Separate workspaces for separate environments (dev/staging/prod). Each workspace has its own agent context, tool registry, and audit scope.

Ready to deploy?

Open source. Self-hosted. Up and running in minutes.

Read the docs →