On June 15, 2026, Anthropic split Claude billing into two distinct tracks: free and paid. The free tier lost automatic access to agentic workflows, and paid tiers moved to a unified credit pool. Anthropic posted the changes on its official pricing page and changelog. The move ended the company’s agent subsidy that had let free users run Claude Code loops without a credit meter. Free users now face a five-hour rolling reset for standard messages. Paid Claude Pro and Max users no longer receive flat-rate, near-unlimited agent access. Instead, every agent run deducts credits based on token volume, tool calls, and context length.
The June 15 date was not a surprise. Anthropic had been signaling tighter free limits since May, when Claude reset its rate limits. But the sudden removal of free agent runs caught many users off guard. A free Claude account could previously trigger a small number of agent tasks per day. After the split, agent access required a paid plan. Anthropic described the change as a necessary step to control the high cost of serving long-horizon agents. The pricing page now lists Claude Pro at $20 per month and Claude Max at $100 per month with credit pools of 1,500 and 7,500 credits respectively.
Who does this hit? Anyone using free Claude for anything beyond casual chat. Students who relied on free Claude to summarize documents saw their context window tighten. Developers who used free Claude Code for agentic coding lost that option. Paid users were not spared. Pro subscribers who previously paid a flat fee for high usage now must monitor a credit balance. Max subscribers gained the largest pool, but overage costs became possible for the first time. This shift fits a broader pattern across the AI industry. Providers are moving from generous free tiers toward metered billing. You can see the wider impact in our free-tier policy report.
Why now? Anthropic’s agent workloads are expensive. Running Claude Code against a large repository can consume millions of tokens per hour. The company had been absorbing that cost for free users. In a competitive market, that was not sustainable. Google and OpenAI also tightened free tiers in May and June. The broader AI price war forced vendors to cut consumer prices while protecting margins on heavy usage. Anthropic chose to keep the $20 Pro price but change what that price buys. The result is a billing split that rewards light chat and punishes heavy automation.
How Do the Top Options Compare?
| Plan | Monthly Price | Standard Message Limits | Agent Access | Credit Pool |
|---|---|---|---|---|
| Claude Free | $0 | 45 messages per 5-hour reset | None | No credits |
| Claude Pro | $20 | Higher base limit | Yes, deducts credits | 1,500 credits/month |
| Claude Max | $100 | Highest base limit | Yes, deducts credits | 7,500 credits/month |
| Claude Team | $25 per user | Shared workspace limits | Yes, team pool | 1,500 credits per seat |
U.S. prices as of June 15, 2026. Agent tasks consume multiple credits based on token length and tool calls. Free limits reset on a rolling five-hour window, not a calendar day. Annual billing for Team lowers the monthly per-seat price to $20.
1. Claude Free , Best for no-cost casual use
Anthropic’s free Claude tier changed sharply on June 15, 2026. The free plan still exists, but it no longer includes agent access. Before the split, a free account could trigger a limited number of agent tasks each day. That ended. Anthropic’s pricing page now lists zero agent credits for free users. The only way to run Claude Code or any multi-step agent is to pay for Claude Pro, Max, or an API key.
The free tier also moved to a five-hour rolling reset. According to Anthropic’s changelog, free users can send 45 standard messages per five hours. The previous daily cap of 100 messages disappeared. For a light user, the new reset can actually be better. You no longer wait until midnight if you exhaust your cap at noon. You wait five hours. But for a user who needs bursts of activity, 45 messages is low. A single long document analysis can consume several messages.
The context window on free Claude tightened for file uploads and long conversations. Anthropic did not publish an exact token count, but users reported that long PDFs now hit context limits sooner. This change pushed free users toward shorter tasks. Our coverage of free-tier limits found that every major AI provider has reduced free access in May and June. Anthropic is not alone. But the removal of free agent runs was one of the sharpest cuts.
Key strengths:
- ✅ Free access to Claude Opus 4.8 in standard mode
- ✅ No credit card required
- ✅ Rolling five-hour reset instead of daily cap
- ✅ Simple chat and light drafting still work
- ✅ Clear paid upgrade path
- ❌ No agent runs after June 15
- ❌ Stricter context limits on long documents
- ❌ 45 messages per five hours can run out fast
Who it’s for: Choose Claude Free if you use Claude occasionally and do not need agentic workflows.
2. Claude Pro , Best for daily professional use
Claude Pro changed from flat-rate access to a credit pool on June 15, 2026. The $20 monthly price did not move. What changed is how usage is counted. Anthropic’s pricing page states that Pro subscribers receive 1,500 credits per month. One standard message costs one credit. Agent tasks cost more, with multipliers based on token length, tool calls, and context window. A long Claude Code session can burn 50 credits or more.
The shift ended the old Pro value proposition. Before June 15, a Pro user could send a high volume of agent requests without watching a meter. That was part of why developers subscribed. After the split, Pro users must budget credits. Many users reacted with frustration on social media and developer forums. The Anthropic agent billing split was one of the most discussed pricing changes of the month.
Pro still includes priority access and fast mode for Claude Opus 4.8. But fast mode also consumes credits, though at a lower multiplier than standard mode. Anthropic released Claude Opus 4.8 at the same price with a cheaper fast mode earlier in June. The combination of a cheaper fast mode and a credit pool means Pro users can stretch credits further if they are willing to use fast mode.
Anthropic added credit top-ups for Pro users. When you exhaust 1,500 credits, you can buy more at a rate of one dollar per 75 credits. This is the first time Pro has had an overage path. It turns a flat subscription into a hybrid model. Compared to OpenAI’s pricing changes, Anthropic kept the base price stable but introduced usage risk. The change means Pro is still good for chat and light coding, but heavy agent users may need Max or the API.
Key strengths:
- ✅ Priority access during peak traffic
- ✅ Credit pool covers standard and agent tasks
- ✅ Access to Claude Opus 4.8 fast mode
- ✅ Same $20 monthly price
- ✅ Clear credit usage dashboard
- ❌ No more flat-rate near-unlimited agent access
- ❌ 1,500 credits can disappear with heavy agent use
- ❌ Overage top-ups add cost
Who it’s for: Choose Claude Pro if you use Claude daily and need reliable agent access without Max pricing.
3. Claude Max , Best for heavy Claude Code and research
Claude Max became the only consumer plan that could sustain serious agentic workloads after June 15, 2026. Anthropic set the Max monthly price at $100 and the credit pool at 7,500 credits. According to the pricing page, Max users get a five times multiplier on fast mode. This means a fast-mode message may cost fewer credits relative to standard mode. Max users also keep the highest rate limits during peak times.
The credit pool replaced the old unlimited-with-fair-use policy. Before the split, Max was the plan for power users who did not want to think about usage. After June 15, every agent run deducts from the 7,500 credit balance. A full day of Claude Code with long context can consume several hundred credits. Anthropic’s changelog documented the token-to-credit conversion. The company said the new model prevents abuse and funds capacity expansion.
Max still represents a better value per credit than Pro. A Max subscriber pays $100 for 7,500 credits, which works out to about 1.33 cents per credit. Pro costs $20 for 1,500 credits, or 1.33 cents per credit as well. The math is nearly identical on paper. The difference is volume and top-up pricing. Max users get a larger base pool and lower priority wait times. The Claude free vs paid comparison shows that Max is the only consumer tier that can run agent loops without immediate exhaustion.
Anthropic added top-ups for Max at a better rate than Pro. The top-up rate for Max starts at one dollar per 100 credits, according to the pricing page. That is cheaper than Pro’s one dollar per 75 credits. Heavy users who exceed 7,500 credits can keep working, but the monthly bill can climb. Some teams are shifting to the Claude API for more predictable pay-as-you-go billing. Others are testing Google’s Gemini pricing cuts as an alternative.
Key strengths:
- ✅ Largest consumer credit pool at 7,500 credits
- ✅ Five times multiplier on fast mode
- ✅ Top-up credits available
- ✅ Highest rate limits
- ✅ Agent runs possible without immediate hard stops
- ❌ $100 monthly price is high
- ❌ Overage costs still possible
- ❌ Credit math still confuses users
Who it’s for: Choose Claude Max if you run long agentic coding sessions or process large volumes of material.
4. Claude Team , Best for small teams sharing a workspace
Claude Team also changed on June 15, 2026. The plan is priced at $25 per user per month for monthly billing. Annual billing lowers the cost to $20 per user per month. Unlike Pro and Max, Team seats share a single workspace and a pooled credit balance. Anthropic’s pricing page states that the Team credit pool is calculated as 1,500 credits per seat per month, added together. A five-person team receives 7,500 credits each month to split.
The shared pool introduces coordination overhead. One heavy agent user can drain the team’s credits and affect colleagues. Anthropic added admin tools to set per-seat limits and alerts. This helps teams avoid surprise exhaustion. Before June 15, Team users had flat-rate access similar to Pro. The shift to credits makes team budgeting more visible but also more restrictive.
Team also includes priority support and a higher context window for shared documents. It does not include the same five times fast-mode multiplier as Max. Team users use credits at the same rate as Pro for standard and agent tasks. For small businesses, the per-seat pricing can look attractive compared to buying Pro for each person. But the shared pool is a major change. Anthropic’s pricing page lists Team as the smallest business plan. Larger enterprises must contact sales.
The Team plan competes directly with Google’s Gemini business tiers and OpenAI’s Team plan. All three vendors have moved toward usage-based billing in 2026. For teams that rely on Claude Code, the new credit pool forces a conversation about cost per agent run.
Key strengths:
- ✅ Shared workspace with centralized billing
- ✅ Team credit pool can be shared across seats
- ✅ Admin controls and usage monitoring
- ✅ Same per-seat price as Pro at $25
- ❌ Per-seat cost adds up
- ❌ Agent credits split among members
- ❌ Not as deep as enterprise features
Who it’s for: Choose Claude Team if you need shared Claude access for multiple people without separate subscriptions.
Frequently Asked Questions
What exactly changed on June 15, 2026?
Anthropic split Claude billing into free and paid tracks. Free users lost automatic agent access and moved to a five-hour reset. Paid Pro and Max users shifted from flat-rate access to monthly credit pools.
Does free Claude still include agent access?
No. Free Claude no longer includes any agent runs. You need a paid plan or an API key to use Claude Code or multi-step agent workflows.
How does the credit pool work for Claude Pro?
Claude Pro users receive 1,500 credits per month. One standard message costs one credit. Agent tasks deduct multiple credits based on token length and tool calls. Fast mode uses fewer credits than standard mode.
How many credits do agent tasks consume?
Agent runs can consume five to fifty credits or more depending on length, context, and tool calls. A long Claude Code session can burn hundreds of credits in a single day.
Can I buy more credits if I run out?
Yes. Anthropic added top-up options. Pro users can buy credits at one dollar per 75 credits. Max users get a better rate of one dollar per 100 credits.
Is Claude Free still worth using?
For occasional chat and light drafting, yes. Free users get 45 messages per five hours. But for serious work, the missing agent access and tighter context limits make paid plans more practical.
What Should You Remember?
- June 15 split: Free Claude lost agent access and moved to a five-hour reset.
- Credit pools: Pro gets 1,500 credits, Max gets 7,500 credits per month.
- Agent tasks cost more: Multipliers can burn 50 credits per run.
- Free tier remains: 45 standard messages per five hours with tighter context.
- Paid price unchanged: Pro stays $20, Max stays $100, but usage caps changed.
- Top-ups added: Pro costs $1 per 75 credits, Max costs $1 per 100 credits.
- Industry shift: Anthropic follows OpenAI and Google in tightening free tiers.
Free AI News is an independent editorial publication. Information about AI pricing, free-tier limits, and features changes frequently and may become outdated. Always verify current details through the vendor’s official pages. Affiliate links may earn a commission at no cost to you, and never affect our reporting.
