Now shipping · AI · OS

Aether OSAether|OS

CodeSoul Labs

Enterprise AI agent operating system — not a chatbot.

Aether OS is for clients who need agents to run real work: ops, support, sales assist, and in-product automation. One control plane for agents, workflows, and decisions. Aether Core is the runtime inside.

See the OS Talk to us

Control plane · demo

Aether Core

Client workspace

Demo

Agents

Tools

Jobs

Planner · job scoped
Tools · connected, allow-listed
Grounding · client docs indexed
Eval suite · running
Human review · optional gate

Control plane

Core

Runtime

Client

Built for you

Safe

Limits first

Product suite

One OS. Agents, workflows, and decisions.

Clients get a system they can operate — not a demo chat box. Aether Core runs the agents. Aether OS is the plane your team watches and steers.

AO · CodeSoul Labs

Aether Core

The runtime clients actually run on

Agent jobs with tools, memory, and checkpoints
Grounded answers on your data — not open-ended theatre
Retries, cancellation, and clear failure modes
Eval and safety as part of the kernel, not a slide

AO · CodeSoul Labs

Control plane

See every agent, workflow, and decision

Live view of agents, models, and connected tools
Workflows you can pause, assign, and review
Logs and traces your ops team can read
One picture for founders, operators, and engineers

AO · CodeSoul Labs

Client embed

Inside your product, under your brand

Assistants and agent UIs in your app or site
APIs so your stack stays yours
Human-in-the-loop where the risk is real
Same craft bar as CodeSoul Labs own products

What clients run on Aether OS

The jobs businesses actually pay for — scoped, logged, and owned in writing before we connect tools.

Ops automation

Internal loops with oversight

In-product assistants

Grounded in your product data

Support agents

Tickets, FAQs, handoff to humans

Sales assist

Research, drafts, follow-ups — not spam

Research & docs

Read, cite, and summarise

Workflow engine

Multi-step jobs with gates

Memory layer

Short-term, grounded, persistent

Model routing

Right model per job, not one ID

Tools & APIs

Only what you allow

Eval & safety

Quality bars before scale

Private deploy

When the brief requires it

Platform capabilities

Multi-agent OS for client work — shared memory, tools, workflows, auth, and observability.

01Agent orchestration — builtin roles and custom agents
02Workflows with human checkpoints
03Task engine — queues, retries, priorities, cancel
04Memory — session, grounded retrieval, persistent records
05Multi-provider model routing (no hardcoded model IDs)
06Plugin / tool layer limited to what you connect
07Auth and role-based operator access
08Logs, traces, and audit trails
09Eval harness before you roll out wide
10Notifications for owners and operators
11Embeddable chat and agent surfaces
12Private deploy path when required

Operator surfaces in one control plane

What your team opens day to day.

Overview

Agents

Workflows

Tasks

Memory

Models

Tools

Inbox

Logs

Eval lab

Analytics

Settings

How a client build works

From your stack to live loops in three steps — no mystery OS.

Scope the jobs

What the agent must do, what it must never do, and where a human stays in the loop. Written before tools connect.

Connect the stack

Data, APIs, and tools Aether Core is allowed to use. Nothing extra. Grounding and policy sit in the OS.

Run from one plane

Loops execute, operators watch, and decisions surface in Aether OS. You scale only after eval clears.

Stack

Next.jsTypeScriptPythonFastAPIPostgresRedisVector searchLLM APIsEval toolingVercel / cloud