Openai Launches Always-on AI Agents
On this page
Reports say OpenAI is launching always-on AI agents, meaning agents that keep running and act without a user prompt for each step. This draft did not check a primary OpenAI source, so it gives no launch dates, feature lists, or pricing. To give any always-on agent access to external APIs, connect it to MCP tools. Each MCP tool exposes one API as a callable tool with a defined schema.
What an always-on agent does
An always-on agent keeps running between user messages. It waits for a trigger and performs work. Then it waits again. A chat session ends when the user leaves, but an always-on agent continues.
Most of these agents share three parts:
- Triggers that start a run, such as a schedule, an incoming event, or a message.
- Tool calls that read data or change state in external systems.
- Persistent state that carries results from one run to the next.
The model decides which tool to call. The host application decides which tools the model can reach.
Why tool access decides whether these agents work
An agent with no tools can only write text. It cannot check an order, post a message, or query a database. Tool access turns the model into a worker.
Each tool call needs three things: credentials, a request shape, and a way to handle failure. A chat user can retry by hand. An unattended agent cannot, so error handling must live in the tool layer or the agent loop.
The Model Context Protocol defines one way for a host to list tools, read their input schemas, and call them. A host that speaks MCP can use any server that speaks MCP. The agent does not need custom code for each API.
Give your agent external APIs through MCP tools
APIbase is an MCP gateway. It exposes third-party APIs as MCP tools. An agent host can list and call those tools. The catalog covers hundreds of providers across many categories, and it changes over time. Check the live listing for current coverage.
A typical setup follows four steps:
- Connect the agent host to the MCP endpoint and confirm the tool list loads.
- Read the input schema for each tool the agent needs.
- Pass arguments that match the schema, and reject any call that does not.
- Parse the JSON response and write the result to the agent state.
Keep the tool list short for each agent. A narrow list reduces wrong calls and makes logs easier to read.
Entry pricing starts at 0 USD, and the highest listed tier is 29.99 USD. Confirm the billing unit on the live pricing page before you budget a continuous agent, because a run that never stops can accumulate calls faster than a person expects.
Limits to set before an agent runs unattended
An always-on agent can repeat a bad action many times before anyone notices. Set limits before the first scheduled run.
- Scope each credential to the smallest set of actions the agent needs.
- Cap calls per hour and per run, and stop the loop when the cap is hit.
- Log every tool call with its arguments, status, and response size.
- Require a human approval step for any write that cannot be undone.
Test the agent with a dry run that logs intended calls but sends none. Compare the log with the expected behavior, then enable writes.
What to check in the OpenAI announcement
Before you design around an OpenAI agent product, read the primary documentation. Confirm three points:
- Whether the agent accepts external MCP servers or uses a closed tool set.
- How the agent authenticates to third-party APIs.
- What the agent charges for continuous runs.
This draft found no public pricing source for the OpenAI product, so it states none. Treat any figure from a secondary report as unverified until the primary page confirms it.
Live pricing — developer
Data source: live tool catalog snapshot, fetched 2026-10-29.
Chart data as a table
| Tool | Price per call | Cache-hit price |
|---|---|---|
| Usage Time Series | $0.0 | $0.0 |
| Per-Tool Usage Breakdown | $0.0 | $0.0 |
| Usage Summary | $0.0 | $0.0 |
| Discover Tools | $0.0 | $0.0 |
| Batch Tool Calls | $0.0 | $0.0 |
| Tool Quality Metrics | $0.0 | $0.0 |
| Tool Quality Rankings | $0.0 | $0.0 |
| List Programming Languages | $0.001 | $0.0001 |
| Check CVE ID Reservation Status | $0.001 | $0.0001 |
| Security Advisories (deps.dev) | $0.001 | $0.0001 |
| Dependency Tree (deps.dev) | $0.001 | $0.0001 |
| Package Info (deps.dev) | $0.001 | $0.0001 |
| Tool | Provider | Price/call | Cache-hit |
|---|---|---|---|
| Usage Time Series | account | $0 | $0 |
| Per-Tool Usage Breakdown | account | $0 | $0 |
| Usage Summary | account | $0 | $0 |
| Discover Tools | apibase | $0 | $0 |
| Batch Tool Calls | platform | $0 | $0 |
| Tool Quality Metrics | platform | $0 | $0 |
| Tool Quality Rankings | platform | $0 | $0 |
| List Programming Languages | judge0 | $0.001 | $0.0001 |
| Check CVE ID Reservation Status | cve-mitre | $0.001 | $0.0001 |
| Security Advisories (deps.dev) | depsdev | $0.001 | $0.0001 |
| Dependency Tree (deps.dev) | depsdev | $0.001 | $0.0001 |
| Package Info (deps.dev) | depsdev | $0.001 | $0.0001 |
Connect via MCP
Data source: gateway design, no measured values.
$ curl -X POST https://apibase.pro/api/v1/tools/account.timeseries/call \
-H "Content-Type: application/json" -d '{"params": {}}'FAQ
Did OpenAI launch always-on AI agents?
Reports describe such a launch. This draft did not check a primary OpenAI source, so it does not confirm dates, features, or availability. Read the official OpenAI announcement and documentation for current details.
What is an always-on AI agent?
An always-on AI agent keeps running between user messages. It starts work from triggers such as schedules or events, calls tools, and keeps state across runs.
Do always-on agents need MCP?
They need some way to reach external APIs. MCP is one open standard that lets a host list and call tools with defined schemas. Whether a specific OpenAI product supports MCP is not confirmed in this draft.
How do I give an agent access to third-party APIs?
Connect the agent host to an MCP gateway such as APIbase, read the input schema for each tool, and pass arguments that match it. Keep the tool list narrow and log every call.
What does APIbase cost for an always-on agent?
Entry pricing starts at 0 USD, and the highest listed tier is 29.99 USD. Check the live pricing page for the billing unit before you budget a continuous run.
Can an always-on agent run without a human watching?
Yes, but it needs hard limits first. Cap calls per run, scope credentials tightly, log every call, and require human approval for irreversible writes.
Is there public pricing for OpenAI's always-on agents?
No public pricing source was found for this draft, so none is stated. Check the official OpenAI pricing page directly.
Recommended next step
- STEP 1/2 Connect your MCP client
- STEP 2/2 Browse the full catalog
Related guides
Published: 2026-10-29Source: live apibase.pro catalog snapshot