Claude Marketplace unifies 2,000 connectors, products, and partners
Anthropic has consolidated its plugins, agent integrations, paid products, and consulting relationships in the Claude Marketplace. The catalog gives enterprises one place to find software that works with Claude, purchase eligible third-party products, and hire firms for large deployments.
The expanded catalog contains more than 2,000 connectors, plugins, agents, products, and service partners. Anthropic launched the marketplace on March 6, 2026, with six partners, including GitLab and Snowflake. Before the expansion, its paid-products section listed 11 partners across cybersecurity, coding, presentation generation, and application deployment.
Three routes through the catalog
The landing page also highlights finance and legal products, including Vanguard Advisor Tools, BlackRock Advisor Center, Addepar, and Paxton Legal Research. Their placement shows Anthropic using the catalog to expand Claude’s reach into regulated, domain-specific workflows.
Committed budgets drive distribution
Enterprises with contractual Anthropic spending commitments can apply eligible funds to third-party marketplace offerings, giving the catalog its main commercial advantage. Anthropic consolidates invoicing and payment processing under the existing customer relationship.
Using committed spend can reduce the work required to purchase a Cursor license, CrowdStrike integration, or another Claude-based product. Customers may avoid establishing a separate billing relationship, although their internal security, legal, and compliance reviews can still apply.
The arrangement also gives partners access to enterprise accounts with approved Anthropic budgets. Anthropic gains a broader product catalog, while vendors gain a purchasing route that can shorten sales cycles.
Anthropic borrows cloud procurement
Cloud marketplaces established this model by allowing customers to apply committed platform spending to third-party software purchased through the same provider. Anthropic adapts that procurement structure to contracted AI spending, turning existing customer commitments into a distribution channel for its ecosystem.
The catalog also expands Claude’s industry coverage without requiring Anthropic to develop every specialized application. Vendors supply expertise in security, legal work, finance, software development, and deployment. Claude remains the model and interaction layer connecting those products.
How builders get listed
Anthropic offers separate public routes for product and agent submissions and connector and plugin publication.
- Products and agents: Apply for commercial distribution through the main marketplace.
- Connectors and plugins: Follow the technical publication process for integrations that extend Claude with external services or capabilities.
A product listing targets centralized enterprise purchasing, while a connector submission targets installation and interoperability. Teams that sell a product and maintain its Claude integration may need to complete both review paths.
Claude Code has its own install flow
Claude Code uses a related plugin system that can package slash commands, agent definitions, reusable skills, Model Context Protocol servers, and lifecycle hooks. MCP servers connect Claude to external tools and data through a standard protocol.
Developers register a plugin marketplace source from Claude Code with the following command, then install plugins from that source’s catalog:
/plugin marketplace add <source>A single installation can add executable hooks, external server configurations, and account-level integrations. Those capabilities make permission and data-flow reviews part of the installation process.
Plugin review starts with permissions
Anthropic requires external plugins to meet its listing standards. Third-party publishers still control the MCP servers, files, and other software bundled with their plugins, leaving runtime security review to the installing team.
- Inspect the plugin manifest and bundled files before installation.
- Identify each MCP server operator and the data sent to that server.
- Review requested credentials, account permissions, hooks, and executable commands.
- Test with least-privileged credentials in a non-production environment.
- Document ownership, update procedures, and incident-response contacts before wider deployment.