# Access and troubleshooting

Resolve provisioning, provider, quota, Discord, and Postiz access problems.


Common Company Brain problems and how to resolve them. If your symptom is
not here, check [How it works](https://erxes.io/docs/company-brain-architecture) for the
expected behavior of each flow.

### Provisioning fails or stays pending

- Provisioning can take several minutes; the page polls the runtime until it
  is serving. Wait for the timeout before retrying.
- A failed deployment keeps the identifier and its reserved plan slot; use
  the retry action. Retry reuses the stored provider and credential mode
  unless you explicitly change them.
- If a subscription provider needs a re-authorization step, the retry card
  shows the provider's credential prompt (or no input for device-code
  providers).

### The chat area shows "Connecting…" for a long time

- The chat surface only mounts once the runtime answers its health check.
  Recent key changes, restarts, or pod recreation keep the overlay up
  briefly by design.
- If it never mounts, retry provisioning from the assistant page, then
  re-open the assistant.

### Provider subscription authorization fails

- Device-code providers (ChatGPT, GitHub Copilot, MiniMax) open a provider
  page with a one-time code; codes expire, so restart the flow if it stalls.
- Claude subscriptions need a setup-token generated by Claude Code on your
  own computer — an API key will not work.
- Plan-key providers (Qwen, Z.AI) need the key issued for the global Coding
  Plan, not the standard pay-as-you-go API key.
- Applying a connection that times out or returns a service error may still
  be in progress; re-open the connection dialog to check the current state
  instead of re-submitting immediately.

### "Limit reached" when creating an assistant

- OpenClaw and Hermes share one assistant entitlement. Business School 3,
  10, and 25 People plans grant 3, 10, and 25 total slots across both kinds.
- Delete an identifier you no longer need to free its slot; identifiers
  created before a failed deployment still occupy a slot until deleted.
- Over-plan assistants use the billing selector; enterprise installations
  are unlimited.

### A team member cannot open an assistant

- Only owners see all identifiers; members must be invited to the specific
  identifier.
- View-only members can open the runtime and see health, but manage actions
  (edit, members, deploy, destroy, transfer) stay with the owner or creator.
- After access changes, the member's open pages pick it up on the next
  check; a reload forces it.

### Discord is connected but the assistant does not respond

- Confirm the channel binding exists for this assistant — bindings are
  assistant- and runtime-scoped, so connecting a channel to one assistant
  does not serve another.
- Confirm the runtime is healthy; the Discord path needs a serving runtime.
- Response controls on the Discord sheet decide which messages get answers;
  check the configured response mode.

### Postiz: "not enabled" or access denied

- On erxes SaaS, an organization must be approved by a platform
  administrator before owners can set up Postiz. Ask your platform contact.
- Uninvited users see guidance only — the workspace URL is never exposed.
- Access is rechecked periodically and on tab focus; removed access unmounts
  the embedded workspace on the next check, not instantly.

### Postiz: owner setup or password issues

- Only the active organization owner can create the tenant workspace; the
  login email is derived server-side and the Postiz password is separate
  from any erxes password.
- The password field clears after submission and is never stored by erxes;
  use Postiz's own recovery if the password is lost.
- The initial Postiz Admin choice is one-time and immutable; if the initial
  admin was deactivated, recovery requires a separately authorized operator
  procedure — there is no automatic owner takeover.

### Postiz: invitations and roles

- Admins invite active erxes users as Admin or Member from the header's
  **Invite team member** sheet; owners have no bypass after initialization.
- The initial admin cannot be removed or demoted, and admins cannot change
  their own role.
- erxes access does not sign anyone into Postiz — normal Postiz login and
  membership still apply, and provider connections are configured in Postiz
  itself.

### CMS content does not reach Postiz

- Sharing is disabled by default; a current Postiz Admin must enable it
  before members can share.
- content_api permissions (assignment, publishing, sharing, author scope,
  language) apply independently — the bridge never bypasses them.
- The bridge verifies matching JWT-derived signing keys on both APIs; if
  `content_api` and `agent_api` were upgraded at different times, envelopes
  fail until both are current.

### Enterprise Postiz enrollment problems

- Enrollment codes are single-use and expire in 15 minutes; request a new
  one after expiry.
- Retries keep the installation identity — a lost response does not create a
  duplicate installation.
- Enrollment is initial-only: once team access is initialized there is no
  self-service re-enrollment or rotation; credential changes need the
  operator recovery procedure.

### A workspace disappeared from the erxes page after expiry

- Expired or revoked access removes only the erxes-side link. Direct
  Postiz sessions, schedules, and workspace data are untouched, and renewal
  restores the same workspace.

### Still stuck?

- [Getting started](https://erxes.io/docs/company-brain-getting-started) — setup walkthrough
- [Setup and Kubernetes](https://erxes.io/docs/company-brain-setup) — operator configuration
- [Social / Postiz](https://erxes.io/docs/company-brain-postiz) — the full Postiz flow
- [Community and support](https://erxes.io/docs/introduction) — Discord and support channels