Agent Self-Provisioning Guide

How an AI agent goes from zero to persistent EU storage in under 10 seconds, with no human in the loop.

1. Register (one-time)

The agent creates its own account. No email verification, no captcha, no credit card.

POST https://api.eustore.dev/v1/auth/register
Content-Type: application/json

{
  "name": "research-agent-7b",
  "email": "agent-7b@mycompany.ai"
}

# Response:
{
  "api_key": "eust_abc123...",
  "customer_id": "cust_...",
  "free_credits": 50,
  "message": "Account created. 50 free credits added (€0.50)."
}

2. Authenticate

POST https://api.eustore.dev/v1/auth/token
Content-Type: application/json

{
  "api_key": "eust_abc123..."
}

# Response:
{
  "token": "eyJ...",
  "expires_in": 86400
}

3. Create a bucket

The agent provisions its own storage. Picks a region closest to its workload.

POST https://api.eustore.dev/v1/storage/buckets
Authorization: Bearer eyJ...
Content-Type: application/json

{
  "name": "agent-7b-memory",
  "region": "eu-central"
}

# Response:
{
  "id": "bucket_...",
  "name": "agent-7b-memory",
  "region": "eu-central",
  "s3_endpoint": "https://fsn1.your-objectstorage.com",
  "created_at": "2026-03-29T07:00:00Z"
}

4. Get S3 credentials

Per-bucket S3 credentials. The agent uses these with any S3 client.

GET https://api.eustore.dev/v1/storage/buckets/{bucket_id}/credentials
Authorization: Bearer eyJ...

# Response:
{
  "access_key": "...",
  "secret_key": "...",
  "endpoint": "https://fsn1.your-objectstorage.com",
  "bucket_name": "agent-7b-memory",
  "region": "eu-central"
}

5. Store and retrieve data

Standard S3 — works with boto3, aws-cli, any S3 library.

# Python (boto3)
import boto3, json

s3 = boto3.client('s3',
    endpoint_url='https://fsn1.your-objectstorage.com',
    aws_access_key_id=creds['access_key'],
    aws_secret_access_key=creds['secret_key'],
    region_name='eu-central'
)

# Store agent memory
memory = {"session": "2026-03-29", "insights": [...], "decisions": [...]}
s3.put_object(
    Bucket='agent-7b-memory',
    Key='memory/2026-03-29.json',
    Body=json.dumps(memory)
)

# Retrieve across sessions
obj = s3.get_object(Bucket='agent-7b-memory', Key='memory/2026-03-29.json')
previous_memory = json.loads(obj['Body'].read())

6. Monitor costs

The agent checks its own spending. No surprise bills.

GET https://api.eustore.dev/v1/billing/balance
Authorization: Bearer eyJ...

# Response:
{
  "credits_remaining": 42,
  "credits_used": 8,
  "estimated_monthly_cost": 0.80,
  "currency": "EUR"
}

7. Top up programmatically

When credits run low, the agent can top up — via crypto (no human approval needed) or Stripe.

# Crypto (instant, no human)
POST https://api.eustore.dev/v1/billing/topup/crypto
Authorization: Bearer eyJ...
Content-Type: application/json

{
  "amount_eur": 10,
  "token": "usdc",
  "chain": "base"
}

# Returns wallet address — agent sends USDC, credits arrive automatically

Real-world patterns

🤖 Long-term memory

# Agent saves learnings after each session
s3.put_object(
    Bucket='agent-memory',
    Key=f'sessions/{session_id}/summary.json',
    Body=json.dumps({
        "date": "2026-03-29",
        "learnings": ["User prefers concise answers", ...],
        "context_for_next_session": {...}
    })
)

# Next session: agent loads all previous learnings
paginator = s3.get_paginator('list_objects_v2')
for page in paginator.paginate(Bucket='agent-memory', Prefix='sessions/'):
    for obj in page.get('Contents', []):
        # Reconstruct memory from all sessions
        ...

📊 Multi-agent shared workspace

# Research agent stores findings
s3.put_object(Bucket='team-workspace', Key='research/topic-a/findings.json', Body=...)

# Writer agent reads research and produces output
findings = s3.get_object(Bucket='team-workspace', Key='research/topic-a/findings.json')

# Reviewer agent reads output and stores feedback
s3.put_object(Bucket='team-workspace', Key='reviews/draft-1/feedback.json', Body=...)

🔄 Checkpoint/resume for long tasks

# Agent saves progress periodically
s3.put_object(
    Bucket='agent-work',
    Key=f'jobs/{job_id}/checkpoint.json',
    Body=json.dumps({"step": 47, "state": current_state, "progress": 0.65})
)

# If interrupted, resume from last checkpoint
try:
    cp = s3.get_object(Bucket='agent-work', Key=f'jobs/{job_id}/checkpoint.json')
    state = json.loads(cp['Body'].read())
    resume_from_step(state['step'])
except s3.exceptions.NoSuchKey:
    start_fresh()