ontbo 1.3.5.260330

Python client for the Ontbo API. Go-to https://www.ontbo.com/.

Make AIs understand your users.

👉 Cognitive Context API for your AI Agents

🌐 Homepage:https://www.ontbo.com

🤖 Developper hub:https://api.ontbo.com

📄 Datasheet: here

💻 Read the full paper:here

Enjoying ONTBO? ⭐️ us to support the project!

Introduction

Ontbo API provides a cognitive context layer for AI agents. It adds long-term memory, persistent context, and reasoning-driven retrieval, enabling AI systems to deliver highly relevant answers and remain consistent over time. Ontbo API plugs into any LLM stack and helps teams build reliable, scalable AI systems.

If your assistants hallucinate, token bills spike, and user context lives in 12 different silos – ONTBO fixes that🔥.

Plug a cognitive layer in front of your LLMs/Agents and stop duct-taping prompt engineering.
Your agents become sharper.

Memory stores. Context understands.

🧩 Main Features

• Context Layer → multi-agent, reasoning-based retrieval that packs only the facts that matter
• Lifecycle Management → facts are versioned, updated, and conflict-resolved automatically
• Autonomous CoT Orchestration → self-directed reasoning → fewer tokens, faster replies, higher accuracy
• Data Governance → white-box by design — full traceability, provenance, and user-level control baked in

🎯 Use Cases

• Personal AI → not memory. Awareness that adapts to you in real time
• Customer Support → not logs. Clear user story that resolves issues on the first touch
• Healthcare → not raw data. Connect all the dots for reliable patient profiles → safer adaptive care
• Developer & Creator Copilots → not autocomplete. Repo + workflow intelligence that guides the next move

🚀 Get started (seriously, 5 min)

1️⃣ Create your account and fetch your api key→ https://api.ontbo.com

2️⃣ Install the lib

pip install ontbo

3️⃣ Your first application in a few lines of code.

from ontbo import Ontbo, SceneMessage

ontbo = Ontbo(token="<YOU_API_KEY_HERE>")
profile = ontbo.create_profile("alice_test")
scene = profile.create_scene("scene_test")

scene.add_messages(
  [SceneMessage(content="Hello, my name is Alice!")], 
  update_now=True)

print(profile.query_facts("What the user's name?"))

💡 Not using Python? See directly our Web API reference → https://api.ontbo.com/api/tests/docs

🤖 Why Ontbo?

The biggest bottleneck for AI agents isn't the model—it's the context. Stop duct-taping prompt engineering and messy RAG pipelines. Ontbo provides a specialized Cognitive Context API that acts as a sophisticated memory layer for your LLMs.

• 📉 Cut Token Waste: Don't resend the same facts. Ontbo optimizes what goes into the prompt.

• 🚫 Kill Hallucinations: Ground your agent with verified, retrieved facts in real-time.

• 🧠 Human-like Memory: Manage complex user histories and cross-session knowledge effortlessly.

• ⚡ Dev-First: Integrates with your existing stack in minutes.

  ⚡ Research Insights - Why you Win ?

  Metric	ONTBO	SoTA
  Recall context	🔥 91%	70%
  Token cost	💸 -95%	Full-data retrieval
  Latency	⚡ 40 ms (P50)	>200 ms
  Model creation speed	🚀 +200%	Baseline

  🔄 4 retrieval modes: best-performance, chain-of-thought, balanced, low-latency
  ✔️ pick precision/latency/cost tradeoffs per request

---

⚡️ Performance Showcase: ChatGPT 5.2 x Ontbo API

Context fragmentation is the silent killer of AI efficiency. See how Ontbo transforms a standard LLM interaction by providing a sophisticated cognitive memory layer.

🎬 Comparative Demo

Feature	ChatGPT (5.2 thinking)	ChatGPT + Ontbo API
Context Retention	Limited / "Forgetful"	Persistent & Cognitive
Retrieval Accuracy	Basic RAG (Messy)	Specialized Memory Layer
Logic Consistency	Dribbles over long sessions	Maintains High-Fidelity Logic

Key Takeaway: While standard models struggle with "duct-taped" prompt engineering, Ontbo ensures your agent actually understands the historical context of the conversation.

🏗 Strategic Benefits (Product & Infra)

• Product → sharper, context-aware responses → higher CSAT & conversions
• Infra → big LLM token savings & lower inference load → measurable cost reductions
• Moat → fine-grained, context-rich personalization with traceable provenance
• Ops → faster model creation & iteration (+200%), predictable scaling, fewer cold starts

This is not merely another vendor-biased benchmark. This performance analysis, comparing ChatGPT 5.2 without Ontbo API and ChatGPT 5.2 with Ontbo API, follows a rigorous methodology formally validated by the Cambridge University Scientific Board in December 2025. Based on 19 items (demographics, cognition, health, finances…). Performance was measured across more than 500 multi-turn conversations, comparing intrinsic LLM memory with an external conversational memory layer.

➡️ This benchmark shows what happens when an external cognitive context layer works alongside leading models, with long-term reasoning reaching up to 93% recall accuracy. ➡️ It’s not a RAG, not memory tools, not a buffer.

📚 Documentation & Support

• API reference docs: https://api.ontbo.com/api/tests/docs
• Community: Discord · X
• Contact: [e-mail]

🤝 License & Contributions

• License: Apache 2.0
• Contributions: PRs welcome. Open issues for bugs/feature requests
• Security: responsible disclosure at [e-mail]

✨ Let’s build something people didn’t think was possible.
🔥 Go break things (and tell us what you build).

If you find this useful, please give us a ⭐ to help others find us!