credential
In one sentence
When the AI needs GitLab / GitHub / Gerrit / DOORS / Jira / Feishu / WeCom / mail / SCM bots, it never sees raw tokens or passwords—it calls credential with intent, and the runtime temporarily injects secrets.
Problems solved
| Without | With credential |
|---|---|
| Tokens leak into prompts / vendors | Tokens never enter the LLM context |
| One token shared across SKILLs scopes up | Scoped per tenant, SKILL, and purpose |
| Revocation requires deleting SKILL files | Disable in console; calls fail immediately |
| No audit trail | Every issuance hits the audit ledger |
Where to configure
Workbench → Integrations → Add credential
- GitLab / GitHub / Gitee / Gerrit
- DOORS Next / Polarion / Jira
- Feishu / WeCom / mail (IMAP / Exchange)
- Custom HTTP (API key / Bearer / Basic)
Industry usage
- Patents: customer IP systems (Patent.com / internal IPDB)
- Automotive: DOORS Next, Polarion, ALM-Octane, Vector vTESTstudio
- Manufacturing: MES, SCADA, ERP (SAP / domestic vendors)
Related
- SCM integrations — per Git host setup
- DOORS integration