Productive MCP Server for Claude and ChatGPT
By Productive
Overview
Productive is the agency management platform that connects sales, projects, resourcing, time tracking, budgeting, and invoicing in one place. This connector gives Claude secure access to your Productive workspace so you can get answers and get work done without switching tabs. Ask about your open tasks, log time against a project, check where a budget stands, or draft a page. Claude works with your live data and respects the same permissions you have in the app. What you can do: Project management. Create and update projects, tasks, and subtasks, including assignees, due dates, statuses, and dependencies. Time tracking. Log, update, and delete time entries, start and stop timers, and manage timesheet submissions. Budgeting and finances. Read and create invoices, expenses, payments, and purchase orders, and check budget and margin status. Sales CRM. Access and update your deals pipeline, companies, contacts, and proposals. Resourcing. View team bookings, time off, and service allocations. People and organization. Look up people, teams, roles, and reporting structure. Docs. Create, edit, and delete pages and comments. Reporting. Query reports across budgets, time, financials, projects, and people. Access and security. Each user connects individually with their own Productive login, and the connection authorizes access only to that account. One connection covers one Productive organization at a time. A separate Sandbox server is available for testing without touching live data.
How to connect
- Claude: Open the Productive listing in the Claude connectors directory and choose Connect. Claude listing Productive exposes 17 tools here.
- ChatGPT: Open the Productive listing in the ChatGPT app directory and choose Connect. ChatGPT listing Productive exposes 17 tools here.
Tools
delete_resource
ChatGPTdescribe_report
ChatGPTdescribe_resource
ChatGPTget_current_person_and_organization
ChatGPTload_resource
ChatGPTitems is a single {type, id} pointer (not the full record). The stub envelope carries output_type: "json" so query_output knows to route through jq. Call query_output(output_id, jq_expression=<jq>) against the handle to read fields. The store envelope is {items: [{id, type, fields}], included?}; fields is keyed by the field's param from describe_resource (e.g. .items[0].fields.title). Relationship values under fields are {type, id} objects (NOT JSON:API .relationships.<name>.data). See query_output for the supported jq subset.load_skills
ChatGPTperform_action
ChatGPTproductive_domain
ChatGPTquery_report
ChatGPTlimit (default 30, max 200) and offset (default 0; must be a multiple of limit). Three call shapes (mutually exclusive): - Fresh query: {report_type, name, group_by?, filters?, formulas?, sort_by?, currency_variant?, include_fields?, display_fields?, report_layout?, chart_type?, limit?, offset?} — issues a new query and returns a query_id you can use to continue it. - Continuation: {query_id, offset, limit?} — the full report config (report_type, name, group_by, filters, formulas, sort_by, currency_variant, include_fields, display_fields, report_layout, chart_type) is pinned from the original call. offset is required — pass the next_offset from the previous response. - Saved report: {report_id, limit?, offset?} — runs the report as saved, so the data matches the report the user sees. The response includes the report's config in query form; adjust it in a fresh query for a modified question. An error carries saved_report_ran: false; a fresh query's results are not this report's data.query_resources
ChatGPTlimit (default 20, max 200) and offset (default 0; must be a multiple of limit). For top-N queries pair sort + limit (e.g., sort=["-created_at"], limit=10). Two call shapes (mutually exclusive): - Fresh query: {resource_type, filters?, sort?, limit?, offset?, include_fields?} — issues a new query and returns a query_id you can use to continue it. - Continuation: {query_id, offset, limit?} — resource_type, filters, sort, and include_fields are pinned from the original call. offset is required — pass the next_offset from the previous response. Use productive_domain to discover all available resource types.read_file_from_url
ChatGPTsearch_help
ChatGPT{query} — returns ranked article titles, URLs, and article content as markdown, with each image marked inline as [image N] instead of its URL; an empty articles array when nothing matches. - Images: {images_for_article_id, image_numbers} — returns each of those placeholders as ready-to-paste image markdown, for an article an earlier search already returned. Only articles from the globally-scoped Help corpus are read — no account-specific data.search_organization_data
ChatGPTsearch_project_data
ChatGPTsearch_records
ChatGPTupdate_resource
ChatGPTcreate_resource
Claudedelete_resource
Claudedescribe_report
Claudedescribe_resource
Claudeget_current_person_and_organization
Claudeload_resource
Claudeload_skills
Claudeperform_action
Claudeproductive_domain
Claudequery_report
Claudequery_resources
Clauderead_file_from_url
Claudesearch_organization_data
Claudesearch_project_data
Claudesearch_records
Claudesearch_resource
Claudeupdate_resource
ClaudeLinks
Similar Apps
Listed on MCP App Store
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/productive-io"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>
Markdown
[](https://mcp-app-store.keenethics-labs.com/app/productive-io)






