tab-sdk 1.0.0

Official Python SDK for TAB Platform — The Verification Layer for AI Agents

tab-sdk

Official Python SDK for TAB Platform — The Verification Layer for AI Agents

Install

pip install tab-sdk

Quick Start

from tab_sdk import TABClient

client = TABClient(api_key="pk_live_xxx")

# Run benchmarks against your agent
result = client.benchmarks.run(
    agent_id="your-agent-uuid",
    benchmark_slugs=["data-exfiltration", "safety-refusal"],
    max_spend=3.00
)
print(f"Score: {result.overall_score}/100")
print(f"Trust Seal: {result.trust_seal_grade}")

Authentication

API Key (recommended)

# Pass directly
client = TABClient(api_key="pk_live_xxx")

# Or set environment variable
# export TAB_API_KEY=pk_live_xxx
client = TABClient()  # auto-reads TAB_API_KEY

Username/Password

client = TABClient(base_url="https://tabverified.ai")
client.auth.login(username="[e-mail]", password="xxx")

Core Operations

Agents

# List your agents
agents = client.agents.list()

# Create an agent
agent = client.agents.create(
    name="My Agent",
    description="Does amazing things",
    agent_type="general",
    tags=["coding", "analysis"]
)

# Get agent details
agent = client.agents.get("agent-uuid")

# Update an agent
agent = client.agents.update("agent-uuid", name="New Name")

# Delete an agent
client.agents.delete("agent-uuid")

# Publish to marketplace
client.agents.publish("agent-uuid")

Benchmarks

# List available benchmarks
benchmarks = client.benchmarks.list()

# List categories
categories = client.benchmarks.categories()

# Run benchmarks
result = client.benchmarks.run(
    agent_id="agent-uuid",
    benchmark_slugs=["swe-bench-pro", "data-exfiltration"],
    model="claude-sonnet-4-20250514",
    max_spend=5.00
)

# Get all your results
results = client.benchmarks.results()

# Get a specific result
result = client.benchmarks.result("result-uuid")

Verification (the Equifax for AI agents)

# Check verification status
v = client.verification.check("agent-uuid")
print(f"Verified: {v.is_verified}")
print(f"Grade: {v.trust_seal_grade}")
print(f"Health: {v.health_score}")

# Get full verification report
report = client.verification.report("agent-uuid")

Marketplace

# Browse verified agents
agents = client.marketplace.list()

# Get agent details with Trust Seal
agent = client.marketplace.get("agent-uuid")

# View leaderboard
leaderboard = client.marketplace.leaderboard()

Credits

# Check balance
balance = client.credits.balance()
print(f"Balance: {balance.balance}")

# View transaction history
transactions = client.credits.transactions()

# Purchase credits (returns Stripe checkout URL)
purchase = client.credits.purchase(amount=20.00)
print(f"Checkout: {purchase.checkout_url}")

Models

# List available models
models = client.models.list()

# List providers
providers = client.models.providers()

Harnesses

# List all harnesses
harnesses = client.harnesses.list()

# Get harnesses for an agent
agent_harnesses = client.harnesses.get_for_agent("agent-uuid")

# Update agent harnesses
client.harnesses.update_for_agent("agent-uuid", harness_ids=["h1", "h2"])

Webhooks

# List webhooks
webhooks = client.webhooks.list()

# Create webhook
webhook = client.webhooks.create(
    url="https://example.com/webhook",
    events=["benchmark.completed", "agent.verified"]
)

# Delete webhook
client.webhooks.delete("webhook-uuid")

Async Usage

import asyncio
from tab_sdk import AsyncTABClient

async def main():
    async with AsyncTABClient(api_key="pk_live_xxx") as client:
        agents = await client.agents.list()
        for agent in agents:
            print(f"{agent.name}: {agent.trust_seal_grade}")

asyncio.run(main())

CI/CD Integration

GitHub Actions

# .github/workflows/agent-verify.yml
name: TAB Agent Verification
on: [push, pull_request]

jobs:
  verify:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'
      - run: pip install tab-sdk
      - name: Run TAB Benchmarks
        env:
          TAB_API_KEY: ${{ secrets.TAB_API_KEY }}
        run: |
          python -c "
          from tab_sdk import TABClient
          import sys
          client = TABClient()
          result = client.benchmarks.run(
              agent_id='${{ vars.TAB_AGENT_ID }}',
              benchmark_slugs=['data-exfiltration', 'safety-refusal', 'prompt-injection'],
              max_spend=3.00
          )
          print(f'TAB Score: {result.overall_score}/100')
          print(f'Trust Seal: {result.trust_seal_grade}')
          if result.overall_score < 70:
              print('FAILED: Agent does not meet minimum verification threshold')
              sys.exit(1)
          print('PASSED: Agent verified by TAB')
          "

GitLab CI

tab-verify:
  image: python:3.11
  script:
    - pip install tab-sdk
    - python verify_agent.py
  variables:
    TAB_API_KEY: $TAB_API_KEY

Error Handling

from tab_sdk import TABClient
from tab_sdk.exceptions import (
    TABError,                    # Base exception
    TABAuthError,                # 401 Unauthorized
    TABForbiddenError,           # 403 Forbidden
    TABNotFoundError,            # 404 Not Found
    TABRateLimitError,           # 429 Too Many Requests
    TABInsufficientCreditsError, # 402 / insufficient credits
    TABValidationError,          # 422 Validation error
    TABServerError,              # 500+ Server error
)

try:
    result = client.benchmarks.run(
        agent_id="uuid",
        benchmark_slugs=["swe-bench"],
        max_spend=1.00
    )
except TABInsufficientCreditsError:
    print("Not enough credits — purchase more at tabverified.ai")
except TABAuthError:
    print("Invalid API key or session expired")
except TABRateLimitError as e:
    print(f"Rate limited — retry after {e.retry_after}s")
except TABError as e:
    print(f"API error: {e}")

All exceptions include status_code, url, and body attributes for debugging.

Type Reference

All request and response types are Pydantic v2 models importable from tab_sdk:

• Agent, AgentCreate, AgentUpdate
• BenchmarkInfo, BenchmarkResult, BenchmarkRunRequest, BenchmarkCategory
• MarketplaceAgent, LeaderboardEntry
• VerificationStatus, VerificationReport
• CreditBalance, CreditTransaction, PurchaseResponse
• ModelInfo, ProviderInfo
• Harness, Webhook
• User, AuthToken

API Reference

Full API documentation: tabverified.ai/static/api-docs.html

License

MIT — see LICENSE