Versanova: Giving AI agents memory and the ability to learn from experience
We are building a memory learning layer for AI agents that lets them remember, adapt, and improve over time — all with a single line of code.
Using our tool, AI agents can store their past experiences, learn from them, share their experiences with other agents, and massively improve reliability and consistency.
Benchmark Results
| Task | Standard | Adaptable-Agent |
|---|---|---|
GAMEOF24 |
12.0% | 99.0% |
SWEBENCH-LITE |
39.2% | 52.2% |
BROWSER-TASK |
60.0% | 100.0% |
BFCL (GLM-4.5) |
76.58% | 86.75% |
Across tasks, Adaptable Agents can deliver massive accuracy gains – all from a single line of integration.
Example Usage
from adaptable_agents import AdaptableOpenAIClient
from openai import OpenAI
# Simply wrap your existing OpenAI client
openai_client = OpenAI() # Uses OPENAI_API_KEY from environment or your existing client
client = AdaptableOpenAIClient(
adaptable_api_key="your-adaptable-api-key",
api_base_url="https://api.adaptable-agents.com",
openai_client=openai_client,
memory_scope_path="my-project/task-name",
)
# Enable adaptable agents (True by default, but explicit for clarity)
client.enable_adaptable_agents = True
# Use it exactly like the OpenAI client - no code changes needed!
response = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "apply arithmetic operators to convert these numbers to 24: 4, 7, 8, 8"}]
)
print(response.choices[0].message.content)
# The client automatically learns from each interaction to improve future responses
Video Demo
Contact Us
Founders
Dr. Gaurav has a PhD in NLP from UCL and previously worked for Yahoo and Amazon.
Dr. Zara has a PhD in machine learning from UCL and has worked for AIG and Booking.
Dr. Amin has a PhD in computer vision from Univ. of Southampton and was founding mlops engineer at Nscale.