Part 2 of 3 · Run
The Autessa AI Native Runtime runs every app you build.
Every app built on Autessa runs on the same set of services: AI agents, agent evaluation, a unified database, screen automation for legacy software, and a private model trained on your data. Each service is versioned, monitored, and audited.
The core services
Four services run every app you build.
AI agents do the work, agent evaluation checks it, the unified database holds the data, and screen automation reaches the software that has no API. Each service is governed and auditable on its own, and all four were designed to run together.
Agents choose from actions you approve, not a hardcoded script.
Each agent works from an explicit set of capabilities and selects the right action for the situation in front of it. Governance, policy gates, and version control are part of every agent from the first run.
Every agent conversation is scored automatically.
Automatic evaluation scores every agent conversation across four dimensions and returns an overall success verdict. You get analytics by version and can drill into any single conversation to see what happened.
One database holds your records, vectors, events, and files.
The unified database combines relational, vector, event-driven, and object storage on PostgreSQL. Knowledge bases give agents retrieval over your own data, and triggers run logic automatically when records are inserted, updated, or changed.
Access to your data is governed inside the database itself, in your Private Cloud.
- Field-level access control lets you decide which users or roles can view, query, or modify each column, so a support agent can search by customer ID without ever seeing the SSN or payment details.
- Masked columns can be stored and searched without being read directly, even by users with database access, which supports identity matching and compliance workflows.
- Every tenant table is encrypted with AES-256 under its own per-tenant key, and you can hold the key that gates model access to your data.
- Reads, writes, and access changes are recorded to the audit trail, so you can reconstruct who touched what, and when.
Agents can work inside software that has no API.
Screen automation lets an agent see what a person sees, understand the context on screen, and interact with the interface directly. That reaches legacy systems and web applications that were never built for integration.
See these four services running on one of your own workflows.
Bring a process that is stuck today. In a 30-minute walkthrough we will show you the agents doing the work, evaluation scoring every conversation, AutessaDB holding the data, and screen automation reaching the systems you thought were out of reach, all inside a Private Cloud you control.
Ready on day one
Your apps can reach people, run on a schedule, and wait for approval without extra tooling.
Beyond the four core services, every deployed app gets the everyday abilities a real workflow needs, from connecting to your existing tools to reaching someone by phone. Each one is governed and audited like everything else.
Connect to the tools you already use
Prebuilt connectors let agents read from and act on the systems your team already works in.
Reach people by phone
Agents can place and take calls, so a workflow can reach a person by voice and not only on a screen.
Run work on a schedule
Recurring work runs on a cadence you set, so agents pick up tasks without anyone starting them by hand.
Route work through queues and approvals
Work moves through managed queues with approval gates and human review at the steps you choose.
Extend with custom code
Built-in scripting covers anything the prebuilt connectors and services do not already handle.
Built for production
Every agent ships ready for production.
Security, versioning, monitoring, and evaluation are part of every agent from day one. You do not add them after something breaks.
Full version control
Every workflow and agent configuration is versioned, so you always know what is running.
Instant rollback
You can roll back to any previous version with no downtime.
Real-time dashboards
Live dashboards show performance and availability across every agent.
Automatic evaluation on every conversation
Every agent conversation is scored as it happens, so quality drift shows up before users notice it.
Encryption and role-based access
Everything you build is encrypted and protected by role-based access control.
Human approval gates
Any step can require a person to approve before the agent continues.
See the full stack running for your use case.
Book a walkthrough and we will show you the model, agents, evaluation, database, and apps working as one system.
FAQ
Questions about the AI Native Runtime
What is the AI Native Runtime, and how does it fit into Autessa?+
Autessa is AI infrastructure for the enterprise: one platform to build AI agents and apps, run them in production, and keep data and models in your control. The AI Native Runtime is the part that runs what you build. Every app runs on the same set of services: a private model trained on your data, AI agents that do the work, automatic evaluation of every conversation, a unified database, and screen automation for software with no API. You build the apps on the Builder Platform, and everything runs inside a Private Cloud you control.
What does agent evaluation measure?+
Automatic evaluation scores every agent conversation across four dimensions: groundedness, task completeness, argument faithfulness, and efficiency. Each conversation gets an overall success verdict, and you can review results by version or drill into a single conversation. This is how you catch quality drift before it reaches users.
What kind of database is it?+
The unified database combines relational, vector, event-driven, and object storage in one system built on PostgreSQL. It adds masked columns, searchable vector columns, and field-level access control, so you do not need a separate vector store, event bus, and object store to support AI workloads.
What makes an agent production-ready?+
Every agent ships with versioning, instant rollback with no downtime, real-time monitoring dashboards, automatic evaluation on every conversation, encryption with role-based access control, and human approval gates on any step you choose. These are part of every deployment from the first run rather than features you add later.
Keep going