Notion Agents can now run a single task for more than 24 hours, and Auto routing claims near-frontier results at about 90% lower cost. Credit tiers and bring-your-own-model options change how you budget and govern agent work.
Key takeaways · AI-distilled
Notion says its upgraded Auto router delivered results similar to Opus 5.5 and Astra across 20+ common use cases at about 90% lower average cost, and re-routes to another model when one stops making progress on a task.
Business and Enterprise seats now include 1,500 Standard credits a month for Auto-routed work and 500 Select credits for chosen models or Custom Agents, kept separate so heavy agent runs cannot exhaust everyday usage.
The sixth-generation Notion agent harnessThe scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.Full definition → runs each session on a queue-backed durable worker loop, which Notion says lets a single task run for more than 24 hours, up from five minutes.
Agents now compact long conversations using providers' native context compactionSummarizing an agent's earlier conversation to free room in the context window so a long session can keep going.Full definition → where possible and offload large tool results from the context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →; Notion reports Custom Agent runs cost about 30% less in testing.
ChatGPT Plus, Pro and Go subscribers can use their plan in the personal Notion Agent instead of Notion credits, and Enterprise customers can bring their own model key in an alpha.
Terms in this piece · Glossary
model routing — Sending each request to a model chosen by the difficulty of the task, rather than using one model for everything.
agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
