Separate product work from AI usage.
Your ManyRows plan covers the workspace and its API limits. Your chosen AI provider manages its own subscriptions, usage charges and data handling.
Give agents the product model, relationships, and evidence behind the work. Ask them to trace a component change, explain release blockers, or prepare a change request. Your team reviews the proposal in ManyRows.
Agents prepare the work. Your team reviews and approves.
Read the types, relationships, and current product records.
Validate values and stage proposed changes for review.
The assigned people review the change and preserve the decision.
ManyRows provides the product model and workflows. Bring a compatible external AI client through MCP; a built-in assistant is not included.
Your ManyRows plan covers the workspace and its API limits. Your chosen AI provider manages its own subscriptions, usage charges and data handling.
Connect an MCP-compatible client to the project endpoint. Check whether your client and subscription support remote MCP and the required authentication method.
Use the client to inspect records, trace BOM relationships or prepare a change request. Access follows the connection’s granted scope, current project permissions and workflow rules.
Start with a concrete question: which assemblies use this part, what blocks this revision, or which changes need review? Agents work through the project’s permissions and existing workflows.
Explore the schema, read records, trace BOM structure, and follow where-used before planning the next step.
Connect an agentDescribe your product and let an agent create record types, fields, and relationships for your team. With schema permissions, it can build on your existing model and validate new definitions.
Explore modellingUse stable business keys, validation, and version checks to connect product records to the systems your team already uses.
Explore the APIStage edits to controlled records in change requests. The team can review proposed work before the approved product record changes.
Explore change controlAsk what blocks the latest product revision. Get document and inspection evidence, source links, and next actions, with gaps in coverage made explicit.
Explore release evidenceOwners and project administrators can follow recorded agent actions and linked reviews. The workspace inbox brings pending, returned, and rejected changes together.
Explore team reviewFollow recorded agent actions to the affected records and reviews. Use the workspace review inbox to find pending decisions, inspect the full change, and request corrections or approve through the existing review policy.
See how changes work
Click to enlarge ↗Start with the agent guide for project identity, effective access, and task entry points. The hosted MCP server exposes permitted tools and leaves human-only decisions out of its executable catalog.
Read the MCP guideUse the JSON Data API for application integrations and automation. Discover capabilities, validate inputs, and make version-aware updates.
Read the API guideNo. Your team can manage parts, BOMs, changes, and quality directly in the web workspace. Connecting AI agents is optional, and you can add them when they fit your workflow.
ManyRows exposes its product model and supported workflows through a hosted MCP server and REST Data API. Agents can discover the schema and work with structured product records using the permissions granted to their connection.
Connect an MCP-compatible client to your project’s hosted MCP endpoint. Use OAuth where supported, or a dedicated API key. The documentation covers endpoints, authentication, capabilities, and supported operations.
API-key agents can propose governed changes, but cannot approve or reject engineering change requests. Human-only operations are excluded from the MCP tool catalog. Every workflow still checks the caller’s authority, including delegated member permissions where supported.
The brief assesses the latest product revision and identifies evidence, blockers, and human next steps. Assembly configuration, as-built state, and shipment eligibility require their own checks.
Yes. Access is bounded by the project and the connection’s permissions. Read-only credentials expose read-safe operations. Schema management requires explicit permission, and workflow rules still apply.
Create a product workspace, then connect an agent if you want help exploring the model or preparing changes for review.