Tenant Deployment

Tenant deployment is the initial setup performed by a 0ai engineer. Once deployment is complete, the FDE begins autonomously observing, learning, and executing. The PM’s role is to share information before deployment and approve observations afterward.

Before Deployment: What PMs Share

Before the 0ai engineer deploys a tenant, the PM shares the following:

Item	Example	Purpose
Organization name	Acme Corporation	Tenant identification and display name
Slack workspace	T0123456789	FDE connection target
Notification channel	#acme-fde	Where results are posted
Dev channel	#acme-fde-dev	For test and error notifications
Approver(s)	PM’s Slack user ID	Target for important decisions
First task to automate	Daily KPI report	Initial scheduled task candidate

Sharing this information with the 0ai engineer via Slack is all that’s needed to start deployment.

Deployment: What the 0ai Engineer Does

The following is handled by the 0ai engineer. PMs do not need to perform any of these operations directly.

1. Tenant Environment Creation

app/orgs/acme/
  skills/
  rules/
  memory/
  config/
  scheduled-tasks.yaml

2. Tenant Registration

.deprex/harness/config/repo-registry.yaml

repos:
  acme:
    display_name: "Acme Corporation"
    slack_workspace_id: "T0123456789"
    dev_channel: "#acme-fde-dev"
    default_notify_channel: "#acme-fde"
    owner_user_ids:
      - "U0123456789"
    content_paths:
      - "app/orgs/acme"

3. Initial Scheduled Task Setup

Based on the business requirements shared by the PM, the first scheduled task is configured.

app/orgs/acme/scheduled-tasks.yaml

tasks:
  - name: daily_kpi_summary
    schedule: "0 9 * * 1-5"
    command: "Review yesterday's key KPIs and post anomalies and action items to Slack"
    env_file: "config/.env"
    description: "Weekday morning KPI summary"

4. Slack Daemon Startup

The FDE’s Slack daemon is started, beginning @mention reception and conversation observation in the designated channels.

After Deployment: What the FDE Starts

Once deployment is complete, the FDE automatically begins:

Automatic Observation

The FDE starts observing conversation patterns in the designated Slack channels:

• What types of tasks are being repeated
• Who is making what kinds of decisions
• What information is frequently referenced

Automatic Initial Memory Construction

The FDE automatically memorizes business knowledge from observations and PM conversations.

@fde Our main products are X and Y. We track monthly ARR and DAU as KPIs

From this conversation, the FDE auto-generates memory/company-overview.md and memory/kpi-definitions.md, then asks the PM to confirm.

Initial Test

After deployment, the 0ai engineer verifies the FDE is operating correctly:

@fde Check Acme's memory and summarize current assumptions

If a response comes back, Slack connection, tenant recognition, and memory recall are all working properly.

The PM’s Post-Deployment Workflow

Week 1: Observation Period

• Receive observation reports from the FDE via Slack
• Review and correct the memory content the FDE has learned
• Test the FDE’s behavior with basic Slack requests

Weeks 2-3: Evaluating Automation Proposals

• Review the tasks the FDE proposes to skillify
• Evaluate quality through Shadow Comparison results
• Approve automation if satisfactory

Week 4+: Normal Operation

• Review scheduled task results and provide feedback as needed
• Approve or reject new automation proposals
• Periodic memory review (the FDE sends reminders)

Important Notes

• Do not fully automate critical tasks immediately after deployment. Start with observation and review
• Credentials (API keys, tokens) are securely configured by the 0ai engineer. PMs do not need to handle them directly
• If the FDE’s output has issues, simply point them out in Slack and the FDE will self-improve