Skip to main content

Anchor Python SDK

Control what your AI agents store. Audit everything.

Installation

pip install anchorai

Quick Start

1. Get Your API Key

Sign up at getanchor.dev to get your API Key.

Important: Copy your API key when you sign up - you won't be able to see it again!

2. Install the SDK

pip install anchorai

3. Initialize and Use

from anchor import Anchor

# Initialize with your API key
anchor = Anchor(api_key="your-api-key")

# Create an agent
agent = anchor.agents.create("support-bot", metadata={"environment": "production"})

# Configure policies
anchor.config.update(agent.id, {
    "policies": {
        "block_pii": True,
        "block_secrets": True,
        "retention_days": 90
    }
})

# Store data (policy-checked, audit-logged)
result = anchor.data.write(agent.id, "user:123:preference", "dark_mode")
print(result.allowed)  # True

# PII is blocked automatically
result = anchor.data.write(agent.id, "user:123:ssn", "123-45-6789")
print(result.allowed)     # False
print(result.blocked_by)  # "policy:block_pii"

# Verify audit chain integrity
verification = anchor.audit.verify(agent.id)
print(verification.valid)  # True

Why Anchor?

  • Policy enforcement: Block PII, secrets, and custom patterns before storage
  • Checkpoints & rollback: Snapshot state, restore if something goes wrong
  • Audit trail: Hash-chained log of every operation, queryable and verifiable
  • Retention policies: Auto-expire data after N days

SDK Structure

The SDK has 5 namespaces:

Namespace Purpose
anchor.agents Agent registry and lifecycle
anchor.config Agent configuration with versioning
anchor.data Governed key-value data storage
anchor.checkpoints State snapshots and rollback
anchor.audit Hash-chained audit trail

Agents

# Create agent
agent = anchor.agents.create("my-agent", metadata={"env": "prod"})

# Get agent
agent = anchor.agents.get(agent.id)

# List agents
agents = anchor.agents.list(status="active")

# Update metadata
agent = anchor.agents.update(agent.id, metadata={"version": "2.0"})

# Suspend/Activate
anchor.agents.suspend(agent.id)
anchor.agents.activate(agent.id)

# Delete
anchor.agents.delete(agent.id)

Configuration

Store any config fields you want. Anchor only enforces the policies section:

# Get current config
config = anchor.config.get(agent.id)

# Update config - store any fields, Anchor enforces `policies`
anchor.config.update(agent.id, {
    "instructions": "You are a helpful assistant",  # Your field
    "model": "gpt-4",                               # Your field
    "policies": {                                   # Anchor enforces this
        "block_pii": True,
        "block_secrets": True,
        "retention_days": 90,
        "retention_by_prefix": {"temp:": 1, "session:": 7}
    }
})

# Config versioning
versions = anchor.config.versions(agent.id)
old_config = anchor.config.get_version(agent.id, "v1")
anchor.config.rollback(agent.id, "v1")

Data Storage

Policy-enforced key-value storage:

# Write data (policy-checked)
result = anchor.data.write(agent.id, "user:123:preference", "dark_mode")
if result.allowed:
    print(f"Stored with audit_id: {result.audit_id}")
else:
    print(f"Blocked by: {result.blocked_by}")

# Write with metadata
result = anchor.data.write(
    agent.id,
    "user:123:topic",
    "billing questions",
    metadata={"source": "conversation", "confidence": 0.9}
)

# Batch write
results = anchor.data.write_batch(agent.id, [
    {"key": "user:123:name", "value": "John"},
    {"key": "user:123:plan", "value": "enterprise"}
])

# Read data
value = anchor.data.read(agent.id, "user:123:preference")

# Read with metadata
entry = anchor.data.read_full(agent.id, "user:123:preference")
print(entry.value, entry.created_at, entry.metadata)

# Search (text similarity matching)
results = anchor.data.search(agent.id, "user preferences", limit=10)
for r in results:
    print(f"{r.key}: {r.value} (similarity: {r.similarity})")

# List keys
keys = anchor.data.list(agent.id, prefix="user:123:")

# Delete
anchor.data.delete(agent.id, "user:123:preference")
anchor.data.delete_prefix(agent.id, "user:123:")

Checkpoints

Snapshot state and rollback if something goes wrong:

# Create checkpoint before risky operation
checkpoint = anchor.checkpoints.create(agent.id, label="pre-migration")

try:
    for item in large_dataset:
        anchor.data.write(agent.id, item.key, item.value)
except Exception:
    # Something went wrong - restore previous state
    result = anchor.checkpoints.restore(agent.id, checkpoint.id)
    print(f"Restored {result.data_keys_restored} keys")

# List checkpoints
checkpoints = anchor.checkpoints.list(agent.id)

# Get/delete checkpoint
cp = anchor.checkpoints.get(agent.id, checkpoint.id)
anchor.checkpoints.delete(agent.id, checkpoint.id)

Audit Trail

Hash-chained audit logging for compliance and debugging:

# Query audit events
events = anchor.audit.query(
    agent.id,
    operations=["data.write", "data.delete"],
    limit=100
)

for event in events:
    print(f"{event.timestamp}: {event.operation} on {event.resource}")
    print(f"  Result: {event.result}")  # "allowed" or "blocked"
    print(f"  Hash: {event.hash}")

# Verify chain integrity (detects tampering)
verification = anchor.audit.verify(agent.id)
print(verification.valid)          # True if chain intact
print(verification.events_checked) # Number of events verified

# Export for compliance
export = anchor.audit.export(agent.id, format="json")
print(export.download_url)

Framework Integrations

Anchor integrates with popular AI frameworks:

pip install anchorai[langchain]  # LangChain
pip install anchorai[crewai]     # CrewAI
pip install anchorai[mem0]       # Mem0
pip install anchorai[all]        # All integrations
# LangChain - Policy-checked memory
from anchor.integrations.langchain import AnchorMemory

memory = AnchorMemory(anchor=anchor, agent_id=agent.id)
# Use with LangChain chains/agents

# CrewAI - Policy-checked shared memory
from anchor.integrations.crewai import AnchorCrewMemory

memory = AnchorCrewMemory(anchor=anchor)
# Use with CrewAI crews

# Mem0 - Policy-checked memory operations
from anchor.integrations.mem0 import AnchorMem0
from mem0 import Memory

wrapped = AnchorMem0(anchor=anchor, agent_id=agent.id, mem0_client=Memory())
result = wrapped.add("User prefers dark mode", user_id="user_123")
print(result.allowed)  # True or False based on policies

Error Handling

from anchor import (
    AnchorError,
    AuthenticationError,
    NotFoundError,
    ValidationError,
    PolicyViolationError,
    RateLimitError
)

try:
    result = anchor.data.write(agent.id, "key", "value")
except PolicyViolationError as e:
    print(f"Blocked: {e.message}")
except AuthenticationError:
    print("Invalid API key")
except NotFoundError:
    print("Agent not found")
except RateLimitError as e:
    print(f"Rate limited. Retry after {e.retry_after}s")

Client Configuration

from anchor import Anchor, ClientConfig

# Simple (just API key)
anchor = Anchor(api_key="your-api-key")

# Full configuration
config = ClientConfig(
    api_key="your-api-key",
    base_url="https://api.getanchor.dev",
    timeout=30.0,
    retry_attempts=3
)
anchor = Anchor(config=config)

Requirements

  • Python 3.8+
  • requests >= 2.28.0

License

Apache 2.0

Metadata

Release files for anchorai 1.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for anchorai 1.1.0
File Size Uploaded Provenance Actions
anchorai-1.1.0.tar.gz 40.8 kB Without provenance Details for anchorai-1.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for anchorai 1.1.0
File Interpreter ABI Platform Actions
anchorai-1.1.0-py3-none-any.whl Python 3 none any Details for anchorai-1.1.0-py3-none-any.whl

Total release size: 77.4 kB

Release files / anchorai-1.1.0.tar.gz

Download URL anchorai-1.1.0.tar.gz
Size 40.8 kB
Tags Source
SHA-256 checksum
How to use checksums
fd35ffd7eef3014560f231a3435874f0d6e3285f432ad30cd02e10818f4429a6
BLAKE2b-256 checksum
How to use checksums
acb77973f16adf9c3e607b6edb3edc54a18621a955ff09a482bd0447a2c97db1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.9

Release files / anchorai-1.1.0-py3-none-any.whl

Download URL anchorai-1.1.0-py3-none-any.whl
Size 36.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
13b88887b1b6d729c976bfedca6d5492da39ea2bbab1c429998ed87682eaa3e4
BLAKE2b-256 checksum
How to use checksums
35718fec7aa66dd28a9ac9d0ff50cc08927a249b15c7e2f2b51d72e55d005fa2
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.9

Security for anchorai 1.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page