1b8bd843d991f81c8d3d7b819fbc7c141ea86cb1
bw's name search is fuzzy: every agent email contains "coo"
(tsgstaff-coo-*), so `bw get totp "coo Cloudron"` matched 10 items and
errored. get_totp now resolves via get_item_id (exact-name filter)
first, mirroring get_item_password.
Added validate-all-logins.py: per-agent fresh-browser-context login
(shared contexts carry session cookies and hide the login form),
asserts password+TOTP round-trip, then verifies every stored API
credential against its system. First full run: 9/10 PASS.
Known failure: vp-compliance stored password does not match the
account ("Incorrect username or password" pre-TOTP) -- enrollment
typed a different value than stored. Needs Cloudron admin reset,
then update_item and re-validate.
Agent Identity Provisioning
Playwright-based automation for provisioning AI agent identities across the TSYS Group stack: Cloudron enrollment (with 2FA), SSO login, API key generation, and Bitwarden credential storage.
Overview
Each AI agent (VP TechOps, VP SecOps, etc.) gets:
- A dedicated Cloudron user (identity root — SSO provisions everywhere)
- TOTP 2FA enrolled and stored in Bitwarden
- API keys generated in Gitea, Discourse, Redmine (stored in Bitwarden)
- All credentials owned by the agent, sourced via
bw-run.sh(no~/.creds/files)
See ~/Q3/agent-identity-bootstrap.md for the full architecture.
Usage
# 1. Create the manifest from the example
cp agents.yaml.example agents.yaml
# Edit: add Cloudron invite links for each agent
# 2. Set BW credentials
export BW_CLIENTID="..."
export BW_CLIENTSECRET="..."
# 3. Build and run
docker compose up --build
# Or run a single agent
docker compose run --rm provision --agent vp-techops
Manifest format
See agents.yaml.example. Each agent defines:
- Cloudron invite link
- Display name
- Priority (Q3 vs Q4)
- System scopes (Redmine projects, Gitea orgs, Discourse categories)
Description
Playwright-based automation for provisioning AI agent identities (Cloudron enrollment, SSO login, API key generation, Bitwarden storage)
274 KiB
Languages
Python
88.4%
Shell
10%
Makefile
0.9%
Dockerfile
0.7%