Agno

CortexDB provides persistent long-term memory for Agno agents, enabling them to learn, remember, and build context across sessions and conversations.

Installation

pip install cortexdb-agno

Quick Start

from agno.agent import Agent
from cortexdb import Cortex
from cortexdb_agno import CortexDBMemory, CortexDBTools

client = Cortex(base_url="http://localhost:3141")

# Create a memory backend
memory = CortexDBMemory(client=client, tenant_id="my-app")

# Create agent tools
toolkit = CortexDBTools(client=client, tenant_id="my-app")

agent = Agent(
    name="Research Assistant",
    memory=memory,
    tools=toolkit.get_tools(),
)

As a Memory Backend

Use CortexDBMemory as Agno's memory provider to give agents persistent recall:

from cortexdb import Cortex
from cortexdb_agno import CortexDBMemory

client = Cortex(base_url="http://localhost:3141")
memory = CortexDBMemory(
    client=client,
    tenant_id="my-app",
    namespace="research",
    top_k=10,
)

# Store memories
memory.add("The quarterly report deadline is March 15th.")
memory.add("Budget for Q2 was approved at $50,000.")

# Search for relevant context
results = memory.search("upcoming deadlines")

# Get formatted context for a prompt
context = memory.get_context("budget information", max_tokens=4096)

# Clear all memories
memory.clear()

Shared vs Per-Agent Memory

# Shared memory — all agents see the same memories
shared = CortexDBMemory(client=client, tenant_id="my-app", namespace="shared")

# Per-agent memory — isolated to each agent
researcher_mem = CortexDBMemory(client=client, tenant_id="my-app", namespace="researcher")
writer_mem = CortexDBMemory(client=client, tenant_id="my-app", namespace="writer")

As Agent Tools

Use CortexDBTools to give agents explicit store, search, and forget capabilities:

from agno.agent import Agent
from cortexdb import Cortex
from cortexdb_agno import CortexDBTools

client = Cortex(base_url="http://localhost:3141")
toolkit = CortexDBTools(client=client, tenant_id="my-app", namespace="tools")

agent = Agent(
    name="Knowledge Manager",
    tools=toolkit.get_tools(),
    instructions="Use CortexDB to store and retrieve important information.",
)

The toolkit provides three tools:

• cortexdb_search — Search for relevant memories using natural language
• cortexdb_store — Save important information for future retrieval
• cortexdb_forget — Remove memories for privacy or data correction

Configuration

| Parameter | Default | Description | |---|---|---| | client | Required | Initialized Cortex client instance | | tenant_id | "default" | Tenant identifier for multi-tenant isolation | | namespace | None | Namespace for organizing memories | | top_k | 5 | Maximum results per memory search |

Complete Example

from agno.agent import Agent
from cortexdb import Cortex
from cortexdb_agno import CortexDBMemory, CortexDBTools

client = Cortex(base_url="http://localhost:3141")

# Shared memory backend
memory = CortexDBMemory(client=client, tenant_id="engineering", namespace="standup")

# Tools for the agent to manage its own memory
toolkit = CortexDBTools(client=client, tenant_id="engineering", namespace="standup")

agent = Agent(
    name="Standup Bot",
    memory=memory,
    tools=toolkit.get_tools(),
    instructions=(
        "You are a standup assistant. Use your memory to track what each "
        "team member is working on. Store updates and recall previous "
        "context when someone asks about project status."
    ),
)

# The agent can now:
# 1. Automatically recall relevant past context via memory
# 2. Explicitly store new information via the store tool
# 3. Search for specific information via the search tool
# 4. Remove outdated info via the forget tool