TherapyCRM

Release notes

New features, improvements and fixes in each TherapyCRM release.

v1.0.21 ·

safer AI-assisted practice workflows

TherapyCRM 1.0.21 makes AI-assisted practice administration safer and easier to correct. New leads and clients stay contact-silent unless staff explicitly request notifications, document responses disclose only necessary creator details, and accidental assessment assignments can be cancelled with a durable audit trail.

New features

  • Practice staff can cancel eligible assessment assignments through supported workflows with confirmation and audit history

Improvements

  • AI-assisted lead and client creation is contact-silent by default, while explicit notification requests remain supported
  • Document and public-share responses now return only the minimum creator and share details needed by the workflow

Bug fixes

  • Fixed accidental outbound notifications during AI-assisted lead or client creation
  • Fixed assessment assignments that could not be safely cancelled after an entry mistake
  • Fixed conflicting assessment start and cancel actions so the recorded status and audit trail remain consistent

v1.0.20 ·

durable AI connections and dependable email tests

TherapyCRM 1.0.20 keeps authorized MCP connectors working across normal periods of inactivity while preserving short-lived access tokens, rotating refresh credentials, tenant isolation, role enforcement, and audit controls. Email template tests now report provider acceptance instead of queue admission.

New features

  • TherapyCRM MCP authorizations can remain connected for up to 30 days of inactivity before reauthorization is required

Improvements

  • Short-lived OAuth access tokens continue to limit exposure while rotating refresh tokens provide a smoother connector experience
  • Email template tests bypass the background queue so administrators receive an immediate, provider-backed result

Bug fixes

  • Fixed MCP connectors requiring reauthorization after the former 15-minute refresh-token window
  • Fixed email template tests reporting success when a message had only entered the local queue and had not yet been accepted by the email provider

v1.0.19 ·

dependable AI-assisted practice workflows

TherapyCRM 1.0.19 strengthens the MCP connector from authorization through day-to-day client, staff, care-plan, scheduling, document, and billing workflows, with concise guidance and clearer recovery from incomplete or invalid information.

New features

  • AI assistants can use expanded machine-readable operation guidance, including required inputs, optional inputs, validation expectations, name-based lookup, and next-step prompts
  • The MCP workspace provides focused interactive UI only where selections or structured review materially improve the workflow

Improvements

  • Name-based resolution supports natural client, staff, service, goal, and scheduling requests without exposing internal identifiers to users
  • MCP responses are shorter, highlight missing required information and validation failures, and suggest the precise correction
  • OAuth consent and completion pages use TherapyCRM branding and provide clear authorize, deny, close-tab, and dashboard actions
  • Appointment and staff-schedule workflows handle realistic partial-input and correction sequences more reliably

Bug fixes

  • Fixed MCP API coverage and query-parameter drift across registered UI-reachable operations
  • Fixed OAuth callback, authorization-submit, connector refresh, and local tunnel integration regressions
  • Fixed appointment time-range errors, scheduling test fixtures, lead conversion branch selection, treatment-schedule full-name search, and deleted client fixture recovery
  • Fixed confusing oversized MCP workspace forms and removed UI cards that duplicated a clearer chat interaction

v1.0.18 ·

reliable OAuth callbacks for AI connectors

TherapyCRM 1.0.18 completes OAuth approval redirects for Claude and other registered MCP clients while retaining a strict, callback-specific Content Security Policy.

New features

  • Claude Desktop can complete TherapyCRM read-and-write OAuth authorization and receive the PKCE authorization code through its registered callback

Improvements

  • The consent page now permits form navigation only to the exact validated callback origin of the registered OAuth client

Bug fixes

  • Fixed the authorization screen remaining on Authorizing after the approval POST completed successfully
  • Fixed Chromium blocking the cross-origin OAuth 302 callback because the global form-action policy allowed only TherapyCRM and Stripe Checkout

v1.0.17 ·

full role-authorized AI assistance

TherapyCRM 1.0.17 lets connected staff agents request read and write access together while preserving the application's existing role, practice, branch, assigned-client, subscription, recent-authentication, and audit enforcement for every operation.

New features

  • Claude and other MCP clients now request TherapyCRM read and write scopes together so the connected agent can use every operation the signed-in staff role permits
  • The authorization screen clearly presents separate view and create-or-update permissions before access is granted

Improvements

  • The OAuth consent button immediately displays an accessible Authorizing state while the secure form submission completes
  • Machine-readable MCP documentation now describes the connector scope policy and confirms that OAuth access never expands TherapyCRM RBAC

Bug fixes

  • Fixed Claude custom connectors receiving a read-only grant even when the signed-in TherapyCRM role permitted mutations
  • Fixed OAuth approval appearing unresponsive during the authorization redirect and token exchange

v1.0.16 ·

a polished, dependable AI connector

TherapyCRM 1.0.16 completes the hosted MCP OAuth experience with branded authorization and recovery pages, reliable continuation through every sign-in path, and stronger release automation so agent connections and deployments remain isolated and verifiable.

New features

  • A TherapyCRM-branded OAuth consent page clearly identifies the requesting AI connector, signed-in account, requested access, and the role and practice protections that remain in force
  • Branded completion and recovery pages explain when it is safe to close the browser tab and provide a direct return-to-dashboard action
  • Claude and other standards-compatible MCP clients can complete the hosted OAuth authorization-code flow through normal login, MFA challenge, MFA enrollment, and temporary-password-change journeys

Improvements

  • OAuth browser errors now render actionable, accessible TherapyCRM pages instead of raw JSON responses
  • Authorization requests return to the server OAuth endpoint after authentication rather than being consumed by the single-page application router
  • Playwright release evidence is isolated across pull-request and staging callers so concurrent workflows cannot read or delete one another's Kubernetes jobs, secrets, pods, or shard results
  • Workflow-tool downloads retry transient network failures and always-run diagnostics retain stable run, shard, and environment identifiers

Bug fixes

  • Fixed the Page not found result that could appear after signing in to authorize an MCP connector
  • Fixed browser-origin compatibility for the narrowly scoped OAuth consent POST while preserving the single-use consent token as the request-forgery boundary
  • Fixed staging qualification failures caused by pull-request and staging Playwright workflows sharing fixed Kubernetes resource names
  • Fixed setup failures that could hide the original CI error behind undefined diagnostic variables