Agents are live: give your workspace a teammate that runs in the cloud

A few weeks ago we said agents were coming: AI with a full harness that watches, reasons and acts on its own. Today they’re here. You can build an agent, hand it tools, deploy it to the cloud, and watch it work — on a schedule, with your tab closed.

Here’s what shipped.

The harness, now with a deploy button

An agent in Nexow is more than a prompt. You assemble it from a harness — model, memory, skills, tools, knowledge and triggers — in a visual builder. That part you may have already seen. What’s new is the last step: Deploy.

Deploying takes a snapshot of your agent and hands it to Nexow’s cloud runtime. From then on it runs on its own heartbeat — every 15 minutes, hourly, daily, whatever cadence you set — without your browser open. You can also poke it manually with Run now whenever you want an off-schedule pass.

Tools are granted, not assumed

An agent only touches what you let it touch. Every tool in its harness is grant-gated: you decide whether it can read your connections, post notifications, write to widgets, and so on. No grant, no access — the agent literally doesn’t see the tool.

That makes it safe to give an agent a real job. A bot tells you BTC dropped 5%; an agent can be told to check the order book and funding rate when it happens, and write up what it found — using exactly the access you signed off on.

Memory between runs

Agents keep memory across runs. What your agent learned on Monday’s pass is still there on Tuesday’s — so it can track a thesis over time, notice changes instead of re-describing snapshots, and skip work it already did.

Every run, on the record

Deployed agents keep a runs timeline: when each run started, whether it finished clean, errored or was skipped, and what happened inside — streamed live as the run executes. Open any run to read the full conversation: what the agent saw, which tools it called, and what it concluded. If an upstream model hiccups, that lands in the history too, labeled as what it is, not a mystery.

You stay in control throughout: pause an agent, resume it, re-deploy an updated harness, or undeploy it entirely. Your plan sets how many agents you can keep deployed at once.

Still private by default

The principle hasn’t moved: the cloud is something you opt into. Building a harness costs nothing and touches no server; deploying is an explicit choice per agent, and privacy mode pauses cloud runs rather than quietly continuing them.

Bots watch. Agents act. Launch Nexow, open the Agents panel, and put your first teammate on the clock.