Your AI assistant and Probara, connected through MCP

Test management your AI agents can operate.

Probara ships a native MCP server: 270+ tools and operations, with up to 14 tools listed by default, so agents plan suites, execute runs, and triage defects with the same permissions, isolation, and audit trail as your team.

Start freeRead the MCP docs

Free plan available · 14-day Pro trial · No credit card required

Works with any MCP client

Connect in one step

Add Probara to your MCP client with one config entry, or to Claude Code with one command.

ConnectPoint any MCP client at your Probara host
JSON
{
  "mcpServers": {
    "probara": {
      "url": "https://app.probara.net/mcp"
    }
  }
}
CLI
# Or add it to Claude Code in one line
claude mcp add --transport http probara https://app.probara.net/mcp

Agents draft the repository, you review it

An agent can create suites and write cases with steps and expected results: the same structures your team edits by hand, ready for review.

Runs executed and recorded, end to end

Agents open runs, submit case results in bulk, and reconcile the run’s case list when the suite changes underneath it.

Failures become defects with the link intact

A failed result turns into a defect in one call, linked to the case and visible on the run, so nothing is lost between tools.

Configuration matrices without the busywork

Agents read a test plan’s combinations, exclude the ones you don’t need, and create a run for the one that needs it, linked to the plan and tagged with its configuration.

270+ tools and operations across every domain

Projects, suites, cases, runs, plans, milestones, defects, environments, custom fields, attachments, imports, members. If the product does it, an agent can do it too. Agents load up to 14 tools by default and reach the rest through search and run tools, so no client loads hundreds of tool definitions.

Browse the full tool reference
  • create_project
  • clone_suite
  • create_test_case
  • replace_test_case_steps
  • create_shared_step
  • promote_shared_step
  • create_custom_field
  • set_test_case_custom_field_values
  • bulk_clone_test_cases
  • create_test_case_review
  • approve_test_case_review
  • merge_test_case_review
  • restore_test_case_revision
  • stage_import
  • create_run
  • bulk_submit_run_result
  • mark_run_case
  • mark_run_case_step
  • reconcile_run_cases
  • bulk_assign_run_cases
  • close_run
  • create_test_plan
  • get_test_plan_combinations
  • set_test_plan_excluded_combinations
  • create_milestone
  • update_milestone
  • create_environment
  • create_configuration_group
  • report_defect_from_result
  • link_defect_case
  • create_defect_jira_issue
  • create_defect_github_issue
  • launch_github_workflow
  • upload_defect_attachment
  • run_pql_query
  • create_saved_query
  • create_dashboard_widget
  • create_traceability_report
  • create_webhook
  • list_audit_events
  • create_invitation
  • list_org_members

Scoped, limited, audited

An agent gets exactly the access you grant, and nothing more.

Organization-scoped
Every call is bound to one organization by its credential. Projects, cases, and runs never cross tenants.
OAuth 2.1 or API tokens
Interactive clients register themselves and a human approves access on a consent screen, with mcp:read / mcp:write scopes. Viewer roles stay read-only. Scripts can send an organization API token as a Bearer credential.
Rate-limited
Per-token and per-organization rate limits keep a runaway agent from overwhelming your workspace.
Audited
Every write an agent performs lands in the organization audit log, the same as any human edit.

Any client that speaks MCP over HTTP connects with no work on our side. The docs carry step-by-step setup guides for the most common ones.

Ready to put agents to work on your testing?

Start free. Connect an MCP client in minutes.

Start freeRead the MCP docs