Anthropic Is Reopening Claude to Third‑Party Agents — But With Strict New Credits

Monthly Agent SDK Credits by Plan

Anthropic’s new credits roughly mirror the price of each subscription tier rather than providing a large automation allowance.

Typical monthly credits reported for individual plans include:

• Claude Pro: $20 per month in Agent SDK credit
• Claude Max 5x: $100 per month in Agent SDK credit
• Claude Max 20x: $200 per month in Agent SDK credit

Once this credit is consumed, programmatic workloads no longer draw from the subscription plan.

What Happens After the Credit Runs Out

After the monthly credit is exhausted, continued usage requires pay‑as‑you‑go billing through the Claude Developer Platform.

That means:

• You must enable billing and use a Claude API key.
• Additional requests are charged at standard API token pricing.
• The costs no longer count toward the subscription’s normal quotas.

This effectively shifts heavy agent workloads into the same pricing structure used by developers building commercial applications on Claude’s API.

Why Anthropic Originally Blocked Tools Like OpenClaw

The new system follows a controversial policy change introduced on April 4, 2026, when Anthropic blocked Claude Pro and Max subscriptions from running external agent frameworks such as OpenClaw using their flat‑rate quotas.

Previously, developers could authenticate third‑party tools with their subscription credentials and run automated workflows without direct API billing. The company ended this approach and required usage‑based pricing instead.

Reports indicated that autonomous agent workloads were consuming far more compute than typical chat sessions, placing significant pressure on infrastructure and engineering resources.

Some analyses suggested individual automation setups could consume thousands of dollars worth of compute per month while paying only a $200 subscription fee.

Why Developers Call the New System a Downgrade

Although the June change restores some access to third‑party agents, many developers argue it doesn’t truly bring back the previous functionality.

Their main criticism: the credit pool is small relative to typical agent workloads, which can continuously generate API calls and token usage.

When the April restriction first took effect, some users reported that switching to API billing increased their costs 10× to 50× compared with running agents under a flat‑rate subscription.

Under the new model, developers say the Agent SDK credit simply acts as a small allowance before the same usage‑based costs apply.

The Bigger Shift in AI Pricing

Anthropic’s move reflects a broader trend across AI platforms: separating consumer subscription usage from high‑volume automated workloads.

Flat‑rate plans work well for interactive chat, but autonomous agents can generate thousands of requests in a short period. By isolating agent workloads behind dedicated credits and pay‑as‑you‑go billing, companies can prevent heavy automation from being subsidized by consumer pricing.

For Claude users, the June 15 update restores technical access to third‑party agents—but it also makes clear that large‑scale automation will increasingly live on metered developer pricing rather than subscription quotas.