
Notion's 2026: autonomous Custom Agents, the credit billing switch on May 4, agents that read your files, 35β50% cheaper runs, and what it costs now.
Notion spent 2026 turning the workspace into an agent platform. Custom Agents (February) run autonomously on schedules or event triggers β not just when you ask. Billing moved to credits on May 4 at $10 per 1,000, agents got 35β50% cheaper to run in April, and the model picker now spans Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, Grok, Kimi and DeepSeek. Full Notion AI is bundled into Business ($20/user/mo) and Enterprise.
| Date | Release | Highlights |
|---|---|---|
| Sept 2025 | Notion 3.0 | Rebuilt around autonomous agents β multi-step work lasting 20+ minutes |
| Jan 20, 2026 | Notion 3.2 | Mobile AI parity Β· AI Meeting Notes transcribe with the screen locked Β· Agent fully functional on mobile |
| Feb 2026 | Custom Agents | Team-wide bots on schedules and triggers |
| Apr 2026 | Notion 3.4 | Custom Agents 35β50% cheaper to run |
| May 4, 2026 | Credits billing | $10 per 1,000 credits Β· monthly Β· no rollover |
| May 13, 2026 | Notion 3.5 | Continued agent and search expansion |
| Jul 1, 2026 | Notion 3.6 | External Agents Β· interactive HTML blocks Β· Workers (code execution) |
Read the table as three phases: capability (3.0 through Custom Agents), economics (April's cost cut, May's credits), and extensibility (3.6's External Agents and Workers). That order matters β Notion built the thing, made it affordable, then opened it.
Notion 3.0's headline claim was agents capable of multi-step work lasting over 20 minutes. That number is the dividing line between "assistant" and "delegate": under a few minutes you supervise, past twenty you hand off and check back.
Custom Agents fire on schedules or events β weekly report drafting, inbox triage, database upkeep. New connectors (Mercury, Mixpanel, Miro, Box, ClickHouse) pull finance, analytics and board data without custom setup.
PPTX, XLSX, DOCX and PDF attachments are agent-readable. The "everything must live in Notion" requirement quietly dropped β which is the single biggest practical unlock for teams that never fully migrated.
Recurring operations β weekly updates, reshaping documents into team formats β become reusable commands rather than re-explained tasks.
Interactive HTML blocks mean an agent's output can be a small working tool inside the page rather than prose about a tool. Combined with Workers' sandboxed code execution, the line between "workspace automation" and "internal app platform" is genuinely blurry.
January's 3.2 made AI Meeting Notes transcribe in the background on iOS and Android with the screen locked, and brought the Agent to full function on mobile. For a product whose core use case is meetings, this was a bigger deal than the version number suggested.
Agent runs can route across Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, Grok, Kimi and DeepSeek. This is a real cost dial: routing repetitive automation to a cheaper model while reserving frontier models for judgment-heavy tasks changes the credit math substantially.
Custom Agents started consuming credits on May 4, 2026: $10 per 1,000 Notion credits, billed monthly, no rollover, on top of a Business or Enterprise plan.
No-rollover is the detail to plan around. Credits are a use-it-or-lose-it monthly budget, not a balance β which rewards steady automation over bursty projects.
Custom Agents became 35β50% cheaper to run in April, with repetitive work like email triage benefiting most. Combined with the May dashboard, this addressed the top complaint from early agent adopters: unpredictable bills.
Opus 4.8 was Anthropic's flagship when the current picker shipped; the Claude 5 family has since arrived. Expect the menu to refresh β don't build documentation around a specific model name.
| Plan | Price | Notion AI |
|---|---|---|
| Free | $0 | Limited |
| Plus | $10/user/mo annual ($12 monthly) | Limited |
| Business | $20/user/mo annual ($24 monthly) | Full AI bundled |
| Enterprise | Custom | Full AI bundled |
Full Notion AI is bundled into Business and Enterprise rather than sold as a separate add-on. For teams that were paying Business plus an AI add-on, this was an effective price cut; for teams on Plus, it's the reason to upgrade.
Plan for seat cost plus credits. A Business team running several scheduled agents will spend meaningfully on credits, and the no-rollover rule means over-provisioning is pure waste. Start low, watch the dashboard for a month, then set the real number.
The 2026 additions aim squarely here. External Agents plus Workers is a low-code internal-tools play as much as a productivity feature β a five-person operations team can build what previously required an engineer.
The compounding advantage remains context: an agent that already holds your notes, tasks, docs and databases starts every job warm (productivity-stack comparison). No standalone assistant can replicate that without the data.
The file-reading capability weakened the main objection β you no longer have to migrate everything first. Evaluate on whether scheduled autonomous work fits your operations, not on whether Notion can host your documents. For hands-on workflows, see the Notion AI productivity guide.
Sources: eesel β Notion AI review 2026 Β· Fello AI β Notion AI pricing 2026 Β· Matthias Frank β Custom Agents tutorial and pricing changes Β· Fayedtion β Notion AI models, features and pricing Β· TechJack β Notion AI credits and real costs
Last updated: July 30, 2026
Anthropic's 2026 in business terms: the $65B Series H at a $965B valuation, confidential IPO filing, $47B run-rate, and enterprise adoption of Claude.
Every major ChatGPT update through July 2026: the GPT-5.5/5.6 transition, Dreaming V3 memory, Lockdown Mode, and what free vs paid gets you now.
Every major Claude update through July 2026: the Claude 5 family, Reflect dashboard, MCP upgrade, Claude Code milestones, pricing and access.
Copilot's 2026: Agent Mode general availability, agentic code review that ships fix PRs, the Copilot CLI, desktop app, and much faster indexing.
Every major Gemini update through July 2026: the 3.5/3.6 Flash generation, Gemini Omni multimodal creation, Nano Banana 2, and the agentic-era strategy.
Midjourney's 2026 in full: V8 alpha to the V8.2 default, native 2K without upscaling, 50% faster and 25% cheaper renders, and 21-second video.
OpenAI's 2026 from the platform side: the GPT-5.x release-and-retire cadence, the Sora shutdown, API changes, and where the company stands mid-year.
Perplexity's browser year: Comet free on every platform since March, iPad and iOS upgrades, Deep Research generating decks, and enterprise controls.