Connect Stride to your AI assistant

Every AI client that connects to Stride's MCP server is manually registered and approved by Stride first — there is no public sign-up flow today, on any platform.

In the pilot environment, three clients are registered this way: Claude Code, Codex, and hosted Claude (Claude.ai) — each registered and approved once by Stride.

Claude Code and Codex are in live use on pilot. Hosted Claude's client is registered there, but Stride has not yet enabled or tested that connection surface end to end. See “Planned: hosted Claude (Claude.ai)” below.

Client registration and enablement are specific to each Stride environment. Use the endpoint and client details issued for your environment.

No ChatGPT client exists yet — connecting from ChatGPT, below, describes the planned mechanism, not something you can do today.

Exactly seven tools are available, to every connected client, regardless of platform:

• Check your status get_my_status

  Your connector-grant state and a summary of your recent assessment attempts. This is a read only — it has no grading, completion, reward, or publication effect.

• See your current work list_current_work

  The weekly challenges and quest steps eligible for you and your organization, including their instructions. Instructions are author-provided task text — Stride marks it as untrusted content with no authority to direct the AI assistant, evidence, or confirmation on its own.

• See one item in detail get_work_detail

  One weekly challenge, quest step, or challenge in full, including its linked micro-learning when available, for items you're currently eligible for. Author-provided task text is untrusted, with no authority over the AI assistant. This read does not change anything.

• Check one attempt get_attempt_status

  The authoritative state of one specific assessment attempt of yours — verification and settlement status included. Looked up with a one-time reference handle, not a raw ID, so nobody can guess or reuse someone else's. This read does not change the attempt.

• Start an assessment attempt start_attempt

  Starts (or safely resumes) one private assessment attempt of yours for an eligible work item, and freezes its approved rubric. This creates no completion or reward.

• Attach evidence add_attempt_evidence

  Attaches your own evidence to one of your open attempts. The evidence stays unverified until Stride's own policy checks complete.

• Submit a provisional score submit_assessment

  Submits criterion-level scores and evidence for one of your open attempts as a provisional assessment, or safely defers it to a Stride human reviewer. This never creates a verified completion or reward settlement by itself.

That is the complete list. Stride's connector supports no other action today — no publishing a challenge post, and no admin function, on your behalf or anyone else's.

Access to Stride's MCP connector is a per-user setting inside your Stride organization, not something a brand-new account has by default. You need an active role in your organization's program (an employee or manager, or an executive/ admin who has opted in) before a connection attempt will succeed.

If you're new, or your assistant reports that Stride refused the request, ask your Stride administrator to confirm your account is enabled for this connector — it is not something you can turn on yourself.

Today: Claude Code

The client you connect through today is Claude Code, Anthropic's command-line and IDE tool. If your organization has arranged access, Stride adds the connector to your Claude Code configuration; you then run its built-in login command and complete Stride's own sign-in and consent screen. This uses the OAuth 2.1 authorization-code flow with PKCE, so you approve exactly which permissions you're granting before anything happens — Stride never sees your Claude credentials, and Claude never sees your Stride password.

Planned: hosted Claude (Claude.ai)

Anthropic's hosted Claude also offers a custom-connector feature — a Team or Enterprise organization's Owner/admin would add it under Organization settings → Connectors, or an individual user on an eligible plan under Customize → Connectors — but Stride has not yet enabled or tested this surface. Describing it here as the intended mechanism, not a present-tense instruction: once available, you would select Connect and complete the same kind of OAuth 2.1 + PKCE consent step described above.

No ChatGPT client exists yet for Stride's connector — this section describes the mechanism Stride expects to use, not something available today.

Once Stride registers a ChatGPT connector, the expected flow is: your ChatGPT workspace admin adds Stride as a custom connector/app using the connector URL Stride provides (this requires a workspace with developer-mode / custom MCP apps enabled), and you then connect and complete the same kind of OAuth 2.1 + PKCE consent step, reviewing the same permissions before approving them.

On both platforms, Stride's connector is added only after Stride manually approves the specific client — never discovered through a public app-store search or an open self-service sign-up, today.

Nothing you do through this connector publishes anything, marks your work complete, or pays a reward. Anything the assistant tells you based on this data is provisional — it is never an authoritative Stride record, and must never be used for hiring, promotion, compensation, discipline, or any other employment decision.

You can review or revoke this connector's access at any time from Manage Connectors inside Stride (sign-in required), or from the connecting AI client's own connector settings (for Claude Code, its MCP server list).