What Backboard.io does

Backboard.io is a memory-native AI infrastructure company headquartered in Ottawa, Canada, founded in April 2025 by Rob Imbeault, co-founder of SaaS unicorn Assent. The company offers developers and enterprises a unified API that delivers persistent memory, stateful context management, and orchestration across 17,000+ large and small language models, directly addressing LLM statelessness and the 'AI amnesia' problem. The platform is built on a proprietary Active Temporal Resonance framework and includes Backboard Studio (recursive code generation), Memory tiers (Light and Pro), AI Router for stateful routing, RAG, Voice API, Image Tool, Adaptive Context Management, and Portable Parallel Stateful Tool Calling, with Python and TypeScript SDKs and OAuth 2.0 / BYOK integrations through OpenRouter and AWS Bedrock.

The company has reported state-of-the-art performance on both major memory benchmarks (90.1% on LoCoMo and 93.4% on LongMemEval, with the latter independently verified by NewMathData) and has been recognized as the first AI platform to lead both metrics simultaneously. Customer deployments target developers and AI/ML teams building agentic systems, with vertical use cases in healthcare (longitudinal patient memory), FinTech (transaction-aware code generation), legal (policy-grounded analysis), and entertainment through a strategic partnership with ESPAT.TV.

Revenue is generated via usage-based infrastructure fees (decoupled from underlying model costs), two memory tiers differentiated by a roughly 10x cost ratio, free state management for MLH participants, and enterprise contracts with volume pricing, dedicated infrastructure, and SLAs. Go-to-market combines product-led growth (self-serve signup, SDKs, documentation), event-driven developer acquisition through the MLH hackathon partnership covering 50+ events and 1M+ developers across 135 countries, and direct enterprise field sales. The company closed a January 2026 pre-seed round led by Mistral Venture Partners with participation from N49P, Garage Capital, Developer Capital, and angel investors, and plans to roughly double its Ottawa team from four to eight-plus.

Name

Backboard.io

Legal name

Backboard.io

Website

https://backboard.io/

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

11–50 employees

Short description

Backboard.io provides developers a unified API delivering persistent memory and stateful context management across 17,000+ LLMs, enabling AI applications in code generation, agentic systems, healthcare, FinTech, legal, and entertainment.

Ownership category

akta.pro rank

Product category

AI Infrastructure

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)

SIC

Services-Prepackaged Software (7372)

akta.pro primary industry

Model Hosting, Serving & Inference Platforms (HDAAACAB)

akta.pro secondary industry

API-First Backend Platforms (BaaS/mBaaS & Serverless Backends) (HDAEAKAH)

Keywords

• AI memory infrastructure
• LLM orchestration platform
• Stateful AI agents
• Model routing API
• AI developer platform

Headquarters

HQ city

Ottawa

HQ country

Canada

HQ region

North America

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Marketing or Sales, Infrastructure, Operations

Revenue model

1. Usage-Based API Access: Consumption-based pricing decoupled from model pricing. Pay for Backboard's infrastructure layer (memory, context management, orchestration) separate from model provider costs. Exposes and controls real drivers of AI spend that token-based pricing hides.
2. Memory Tiers (Light/Pro): Two memory versions with different cost-accuracy tradeoffs. Memory Light at ~1/10th cost of Pro. Memory Pro for mission-critical use cases requiring highest accuracy.
3. Enterprise Volume Contracts: Volume pricing, dedicated infrastructure, SLAs, and custom contracts for enterprise customers with dedicated account management.

Pricing tiers

Go-to-market motion3 records

Distribution channels4 records

Marketing channels7 records

Core offering

Backboard.io provides a developer-first AI infrastructure platform that delivers persistent memory, adaptive context management, and orchestration for AI applications across 17,000+ LLMs through a single unified API. The platform includes persistent memory tiers (Light and Pro), stateful model routing with BYOK support, RAG workflows, parallel stateful tool calling, and developer tooling (Backboard Studio IDE and R-CLI) for recursive code generation. It is sold to developers, software teams, and enterprises building stateful AI agents and applications, with usage-based API pricing and enterprise volume contracts.

Product overview

Backboard.io is a developer-first AI infrastructure platform positioned as the 'memory layer for AI.' It is a unified platform-plus-modules architecture offering a single API that provides persistent memory, model routing across 17,000+ LLMs, multi-agent coordination, RAG workflows, long-term context retrieval, and tool calling. The core platform is the Backboard API with a unified interaction flow. Key products include: Backboard Studio (recursive CLI and IDE for code generation), R-CLI (terminal code generation tool), AI Router/Stateful Routing (stateful routing across models with BYOK support), RAG (retrieval-augmented generation), Memory (Lite and Pro tiers for persistent memory), Voice API (STT/TTS via OpenAI and ElevenLabs), Image Tool (image generation within threads), Adaptive Context Management (automatic context window handling), Portable Parallel Stateful Tool Calling (concurrent multi-tool execution), Custom Memory Orchestration (configurable per-assistant memory rules), Manual Memory Search (API for inspecting agent knowledge), Web-Search (real-time search with memory), and Backboard SSO (OAuth 2.0 authentication). The platform supports BYOK for OpenRouter, AWS Bedrock, and major embedding providers (OpenAI, Google, Cohere). Backboard achieved #1 ranking on both LoCoMo (90.1%) and LongMemEval (93.4%) memory benchmarks.

Differentiator

Problem solved

Functional benefit

Products and services

• Backboard Studio A development environment combining a recursive CLI and IDE for code generation using frontier models. R-CLI decomposes goals into steps, runs them, verifies results, and recurses until completion. Studio generates, reviews, connects, and ships code from prompt to production, achieving up to 30% fewer tokens and 5-10x more output efficiency than standard API calls. Built for developers building AI-powered applications.
• R-CLI Recursive command-line tool for terminal-based code generation that decomposes a goal into steps, executes them, verifies the results, and recurses until the task is complete. It is the entry point into Backboard Studio and is installed via a curl one-liner. Built for developers working in their terminal.
• AI Router (Stateful Routing) An AI routing product that enables stateful routing across 17,000+ LLMs including OpenAI, Anthropic, Google, Cohere, AWS Bedrock, and OpenRouter. It manages conversation state when switching between models with different context window sizes, with BYOK support so developers can use their existing API keys. For developers and enterprises needing to route AI calls across multiple providers without losing state.
• RAG (Retrieval-Augmented Generation) Retrieval-Augmented Generation product integrated into the unified Backboard API. Supports configurable embedding models (OpenAI, Google, Cohere) with vector dimensions from 256 to 3072, document uploads (up to 5M tokens per file), and multimodal RAG for context augmentation. For teams building AI applications that need grounded, knowledge-based responses.
• Memory (Light and Pro Tiers) Persistent memory product offered in two tiers: Memory Lite (~1/10th the cost of Pro, message-level memory for affordability) and Memory Pro (highest accuracy and depth for mission-critical workflows). Provides portable, configurable memory across 17,000+ models with adaptive context management. For developers and enterprises that need AI systems to remember, retrieve, and build on past interactions.
• Voice API First-class voice API providing Speech-to-Text (OpenAI Whisper, ElevenLabs Scribe) and Text-to-Speech (OpenAI TTS, ElevenLabs) capabilities through three operational modes: audio-in/text-out, text-in/audio-out, and full audio-in/audio-out. Supports streaming via SSE with provider options. For developers building voice-enabled AI agents.
• Image Tool Built-in streaming image generation tool for Backboard threads that enables text-to-image generation and image-to-image edits within conversations, using models available via OpenRouter and Google providers. Document IDs enable follow-up edits. For developers building multimodal AI agents.
• Web-Search Real-time web search capability combined with persistent memory for up-to-date responses. Integrated into the unified Backboard API as a RAG workflow building block. For developers building AI agents that need current, factual information grounded in retrieved web content.

Quantifiable outcome

• 90.1% accuracy on LoCoMo benchmark vs competitors at 66-75%
• +4 more outcomes

Named customers4 records

Segments6 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration9 records

AI capability13 records

Feature10 records

Partnerships

Three partnerships are on record, tiered core.

• Major League Hacking (MLH)coreGTM or Marketing Partner · 9 April 202612-month promotional partnership establishing Backboard as the primary AI memory layer for MLH's global hackathon league. Covers 50+ hackathon events worldwide, 3 Global Hack Weeks, and DEV challenges. 'Best Use of Backboard.io' prize category. Free state management for life tier for MLH participants ($5 dev credits). Reaches 1M+ developers across 135 countries.
• ESPAT.TVcoreStrategic or Co-development Partner · 3 February 2026Strategic partnership making Backboard.io the foundational infrastructure layer for ESPAT.TV's next-generation AI-powered tools in entertainment, gaming, and creative industries. Grounded in shared memory-first architecture philosophy. Initial focus on targeted pilots and infrastructure integration with potential expansion across entertainment domains.
• OpenRoutercoreTechnology or Integration · 2 February 2026BYOK (Bring Your Own Key) functionality for OpenRouter users. Developers can add persistent memory and state to existing OpenRouter integrations without refactoring. Existing API routing remains intact with zero disruption. Enables stateful threads, durable memory, session continuity, and integrated RAG capabilities.

Scale indicators8 records

Recent moves6 records

Expansion highlights9 records

Direct peers

• Mem0: Direct competitor in the AI memory layer category. Mem0 provides persistent memory for LLM applications and is one of the primary benchmarks Backboard compares against (66.88% on LoCoMo vs Backboard's 90.1%).
• Zep: Direct competitor offering long-term memory and context engineering for AI agents. Zep scored 75.14% on LoCoMo and is a key benchmark reference for Backboard's memory performance claims.
• LangMem: LangChain's long-term memory module for LLM applications. Scored 58.1% on LoCoMo and represents the integrated memory approach against which Backboard's standalone memory layer competes.
• Supermemory: Direct competitor building infinite memory infrastructure for AI applications and agents. Targets similar developer and enterprise use cases for persistent context across conversations.
• Letta: Direct competitor (formerly MemGPT) building stateful agent infrastructure with persistent memory. Targets the same developer-first, memory-for-AI-agents use case as Backboard.

Broad incumbents

• OpenAI: Broader incumbent that already ships native memory features in ChatGPT and could expand memory capabilities across its API, directly competing with or obviating Backboard's memory layer.
• LangChain: Broader incumbent in the LLM application framework category with LangGraph, LangSmith, and LangMem offerings. Competes with Backboard at the orchestration and agent infrastructure layer.

Emerging players

• Cognee: Emerging open-source memory and knowledge layer for AI applications using ontologies and graph structures. Targets similar use cases for stateful, knowledge-aware LLM applications.
• Pinecone: Adjacent competitor in the vector database / RAG infrastructure space. While focused on retrieval rather than conversational memory, Pinecone increasingly overlaps with Backboard's RAG and long-term memory offerings.

Others

• OpenRouter: Both a partner (via BYOK integration) and adjacent infrastructure player providing unified LLM routing across hundreds of models. OpenRouter's roadmap into memory/state could either deepen partnership or become a direct competitor to Backboard's router layer.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat5 records

Key risks5 records

Key highlights6 records

Customer concentration

• Xx.com/BackboardIO
• GitHubgithub.com/backboard-io
• LinkedInlinkedin.com/company/backboard-io

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds1 record

Investors4 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Backboard.io