What's New in Ingram Cloud: Self-Scheduling Agents & More
Since we announced Ingram Cloud, we've been shipping steadily. This update rounds up four changes that share a theme: agents that act on their own schedule, meet your developers in whatever stack they already use, and feel native in the chat apps your users live in.
Agents that schedule themselves
An agent can now set up its own recurring work — from inside a conversation, with no code on your side. When a user says "remind me every Monday at 9" or "wake me up tomorrow at 8am," the agent turns that into a real schedule it owns: a cron entry that runs it again at the right time and lets it reach back out.
This rides a new built-in tool, manage_schedules, that an agent can use to
create, list, and delete its own schedules. Two things make it safe to hand
to an agent:
- It's scoped to itself. An agent can only ever see or change its own schedules — never another user's. It holds no privileged credential.
- It thinks in the user's timezone. Scheduling is only useful if "8am" means 8am where the user is. The tool requires a timezone and prompts the agent to ask the user for theirs, and the underlying scheduler is daylight-saving aware — a daily 8am job stays at 8am across the spring and autumn clock changes.
It's off by default. Turn it on per agent when you want self-scheduling, and the cost stays governed by the same per-agent and per-user budgets you already set. See the tools documentation to enable it.
Never stall on an empty wallet: auto-reload
Long-running agents shouldn't grind to a halt mid-task because a prepaid balance ran dry. Auto-reload keeps the lights on: when your balance falls below a threshold you choose, we charge your saved card for a top-up amount you choose — off-session, no one in the loop.
It's opt-in and fully under your control: pick the threshold and the refill amount, or leave it off and top up manually. Either way, an agent in the middle of a job won't be the thing that discovers the wallet is empty.
Drop an agent into your stack: new adapters for eve and Flue
Ingram Cloud is standards-first — every agent is reachable over an OpenAI-compatible API, so most tools already "just work." On top of that base we publish thin adapters so an Ingram agent slots into the framework your team already uses, as a first-class citizen rather than a raw HTTP call.
Joining @ingram-tech/ai-sdk-adapter
(our Vercel AI SDK integration), two new packages:
@ingram-tech/eve-ingram-cloud—ingramCloudModel()runs an Ingram agent as the model behind an eve agent, anddefineIngramMcpConnection()wires an Ingram MCP deployment in as eve tools.@ingram-tech/flue-ingram-cloud—registerIngramCloud()registers an Ingram agent as a model provider in Flue, andconnectIngramMcp()brings an Ingram MCP deployment in as Flue tools.
Both expose the same approval helpers, so tool calls that need a human's sign-off resolve through one consistent channel regardless of framework.
Quick replies, the native way, on Telegram and WhatsApp
The fastest answer is often a tap, not a typed sentence. Agents can now offer quick-reply chips — Yes / No, a short menu, a next step — and they render as the platform's own tappable buttons: a reply keyboard on Telegram, native reply buttons on WhatsApp. Tapping one sends it straight back as the user's next message.
The agent declares the options once with a suggest_replies tool; each channel
lowers them to its native affordance, and channels without one fall back to clean
text automatically. It lights up on its own wherever the agent is deployed to a
channel that supports chips — nothing to configure.
Try it
Ingram Cloud runs today. If you're building agents and want them to run on their own schedule, stay funded, fit your stack, and feel native in chat — you can start now.