Anthropic adds up to $500 a month of pooled API credit to Claude Team plans

The per-seat rates come from the help center's worked example. Three Standard seats and two Premium seats receive $260 a month, and adding a third Premium seat raises the next cycle to $360 ^[5][6]. That implies $100 per Premium seat and $20 per Standard seat ^[20]. The $500 ceiling arrives at five Premium seats or 25 Standard ones ^[21]. A team of 50 Standard seats would receive the same $500, or $10 a seat ^[22]. One Max 20x subscriber gets $200 a cycle, 40% of what a whole team can claim ^[17][23].

Where the money lands matters more than how much there is. Only a Team's Primary Owner or Owner can claim it, by linking one Claude Console organization, either an existing one or a new one ^[9]. The credits are rolling out over a few days, and new subscribers wait seven days on an eligible plan before claiming ^[18][7]. Everyone with an API key in the linked organization draws from the balance, and the monthly credit is spent before anything the organization has bought ^[11][10]. Linked to an organization already carrying paid API traffic, the $500 is used up by that traffic first and shows up as a smaller invoice ^[10][11]. I'd expect a separate organization to suit teams that want the credit to pay for new internal agents, with Console workspace spend limits dividing it by project or teammate ^[12].

Running out has two readings. With no other credits in the organization, API requests stop until the next deposit, and nothing is charged to the Claude plan ^[13]. Organizations with purchased credits or auto-reload keep running on those, and customers invoiced through Anthropic sales are billed as usual beyond the credit ^[14][15]. In an organization with no other credits, a pilot cannot overspend, and no card is needed to start ^[13][8]. For a team whose agent has become part of someone's working day, the same rule produces an outage partway through a month. Unused credit expires at the end of each cycle, so a quiet month banks nothing for a busy one ^[16].

The help center does not say how many requests $500 buys. Whether a given internal agent fits inside the pool depends on which Claude model it calls and how often ^[3].

The scope pulls the build onto Anthropic's own platform. The credit covers the Claude API, the Console Playground, Claude Managed Agents and the Claude Agent SDK ^[3]. It does not reach interactive Claude Code, or Claude bought through Amazon Bedrock, Google Cloud Vertex AI or Microsoft Foundry ^[4][19]. A company whose Claude spend runs through one of those clouds can use the credit only in a direct Console organization. This quarter, that choice adds no spend beyond the subscription and needs no card ^[8]. Next quarter, any tool that outgrows $500 there forces a choice between buying credit on Anthropic's platform and moving the tool to the cloud account the company already pays through.