Code for America and Anthropic announced a partnership on May 8 at the 2026 Code for America Summit in Chicago to build Claude-powered AI tools that help government caseworkers administer public benefits. The first product, SNAP Policy Navigator, is built on Anthropic’s Model Context Protocol (MCP) and gives caseworkers real-time, source-cited answers to Supplemental Nutrition Assistance Program policy questions that vary across federal, state, and county jurisdictions.
How SNAP Policy Navigator Works
SNAP caseworkers interpret eligibility rules that change frequently and differ across jurisdictions. The Policy Navigator is designed to return case-specific answers drawn from current, verified policy documents rather than requiring manual searches across multiple documentation sources, according to Nextgov.
A caseworker inputs a policy question, such as how a client’s change in income or a new federal directive affects their benefits. The tool returns a response in plain language with cited sources and suggested next steps. Critically, the tool provides policy clarity, not eligibility decisions. “The decision stays with the caseworker,” Jana Rhyu, VP of Product at Code for America, told attendees at the Summit, as reported by Nextgov.
The tool is built on MCP, Anthropic’s open standard for secure, bidirectional communication between AI applications and data sources. In this case, MCP connects Claude to SNAP policy documents and ensures every response is traceable to a specific policy source, according to EdTech Innovation Hub.
Why Now: The HR 1 Policy Shock
The deployment comes as SNAP administration faces acute pressure. President Trump’s “Big, Beautiful Bill” (HR 1), signed last July, expanded work requirements, shifted administrative costs to states based on SNAP payment error rates, and required cost-neutral recalculations of the Thrifty Food Plan. SNAP participation has declined by more than 3 million people across 36 states as of January, with further reductions expected once the new rules are fully implemented, according to the Center on Budget and Policy Priorities via Nextgov.
“Policy is constantly changing, and the complexities of policy implementation are immense, which places an even bigger burden on the caseworkers,” Rhyu said, as reported by Nextgov.
Michael Lai, who leads state and local government AI at Anthropic, pointed to one former caseworker who described their job as “an email inbox that’s always full, where each one requires care and attention, but you’re constantly getting interrupted,” according to Nextgov.
Broader Suite Planned
The partnership extends beyond SNAP. Code for America and Anthropic plan to develop reusable Claude integrations adaptable across states and counties, covering eligibility document review, policy queries, and drafting plain-language communications to benefit recipients, according to EdTech Innovation Hub.
Code for America worked in 27 states and Washington, D.C. in 2025, helping 7 million people access $22 billion in benefits. That existing infrastructure gives the partnership an immediate pathway to pilot and scale tools in live government environments.
Anthropic’s Public-Sector Play
Elizabeth Kelly, Head of Beneficial Deployments at Anthropic, framed the partnership as Anthropic’s approach to placing AI “where human need is deepest,” according to EdTech Innovation Hub. The move contrasts with OpenAI’s enterprise-services focus through its newly launched Deployment Company and Daybreak cybersecurity initiative. While OpenAI pursues Fortune 500 and government cybersecurity contracts, Anthropic is targeting civic infrastructure where agent utility can be measured in processing times and error rates, not just revenue.
For teams building MCP-based agent integrations, the SNAP Policy Navigator is one of the first production deployments where MCP’s source-tracing capability directly addresses a regulatory compliance requirement: every AI-generated answer traceable to an authoritative policy document.