Shodh Memory
Zero-LLM persistent memory for AI agents — deterministic, auditable, offline.
Shodh Memory is a standout open-source option for developers building autonomous agents or robotics systems that need deterministic, offline memory. It outpaces cloud-dependent alternatives in latency and cost, but non-technical users may find the CLI/API approach challenging. If your project demands privacy and no LLM in the loop, it's hard to beat.
Verified 4d ago · liveness 70/100 · cite: rightaichoice.com/tools/shodh-memory
- Developers building autonomous AI agents needing deterministic memory
- Robotics engineers requiring offline persistent memory for ROS2/Zenoh platforms
- Edge/IoT developers working on Raspberry Pi or air-gapped systems
- Researchers exploring cognitive architectures and Hebbian learning
- Users needing cloud-synced or team-shared memory out of the box
- Enterprise teams requiring SSO, RBAC, or hosted management UI
- Non-technical users seeking a GUI-only memory tool
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Shodh Memory if you need a hosted, team-shared memory service with SSO/RBAC or a GUI, or if you rely on LLM-generated summaries in your memory loop.
Models auto-download on first use (~37MB), so you need a one-time internet connection for initial setup.
Shodh Memory is free and open source (Apache 2.0), making it a cost-effective choice for developers comfortable running their own memory infrastructure. Unlike cloud memory services like mem0 or Zep that charge per memory operation or per token, Shodh has zero ongoing costs. Enterprise support is available for teams needing help, but most users can get everything from the open-source version.
In short
Shodh Memory — Zero-LLM persistent memory for AI agents — deterministic, auditable, offline. Best for Developers building autonomous AI agents needing deterministic memory, Robotics engineers requiring offline persistent memory for ROS2/Zenoh platforms, Edge/IoT developers working on Raspberry Pi or air-gapped systems. Free to use.
What's new in Shodh Memory
Checked 2 days agoAcross the latest 5 updates: 5 news mentions.
Causal Retrieval: The Memory Problem Vector Search Can't Solve
Explains how Shodh's knowledge graph enables causal retrieval through backward graph walks, solving a problem vector search cannot.
Language Models Are Few-Shot Learners — But They're Amnesiacs
Argues that persistent memory is key to making few-shot learning persist without an LLM in the loop.
Types of Memory in AI: From Working Memory to Long-Term Potentiation
Publishes a comprehensive guide on memory types from neuroscience to AI systems, grounding Shodh's architecture.
Cognitive Architectures for AI Agents: From ACT-R to Modern Memory Systems
Maps cognitive science models to AI agent memory, highlighting the 'limited memory AI' problem.
Graph Databases for AI Memory: Why Your Agent Needs a Knowledge Graph
Argues for knowledge graphs enabling causal reasoning and spreading activation in AI memory systems.
What people actually say about Shodh Memory — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
5 mentions across 2 sources (Hacker News, GitHub) · researched Jul 3, 2026.
- +Zero LLM calls for memory operations—fast and cost-free.
- +Fully offline; data never leaves the machine.
- +Single ~30MB Rust binary, no Docker or dependencies.
- +Hebbian learning and decay curves mimic biological memory.
- +Sub-microsecond graph lookups; 34–58ms semantic search.
- −Very early stage—only 227 GitHub stars and 9 issues.
- −Sparse documentation for advanced features.
- −No cloud sync or collaborative shared memory.
- −Limited community examples and third-party tutorials.
- −Integrations beyond listed MCP/frameworks are thin.
- • No hidden costs—fully free and open source.
- • Self-hosting may incur infrastructure costs for server mode (Docker optional).
Viability Score
How well maintained and how widely used is Shodh Memory? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- Zero LLM calls in memory loop
- Hebbian learning with decay curves
- Knowledge graph with temporal indices
- Hybrid ranking (vector + graph + time)
- Sub-microsecond graph lookups
- Semantic search in 34–58ms
- Blind spot detection and stale zone detection
- Causal retrieval via backward graph walk
- Local-only operation — data never leaves the machine
- Runs on Raspberry Pi Zero, Jetson, air-gapped systems
- REST API with health endpoint
- 37 MCP tools for agent integration
- Client libraries for npm, PyPI, crates.io
- Docker image for server mode
- Offline-first architecture
About Shodh Memory
Shodh Memory is a local-first persistent memory system for AI agents that removes LLM calls from the memory loop entirely. Instead of using an LLM to store, summarize, or manage memories—which is slow, expensive, and non-deterministic—Shodh uses Hebbian learning rules and decay curves. Memories strengthen when co-accessed, fade naturally over time, and wire together like biological synapses. Every memory operation is deterministic, auditable, and completes in microseconds, with all data staying on your machine. Built for developers building autonomous agents, robotics systems, edge devices, and air-gapped environments, Shodh runs as a single ~30MB Rust binary with no Docker or external dependencies. It combines a knowledge graph with temporal indices and hybrid ranking, enabling causal retrieval and blind spot detection that vector-only memory stores can't achieve. Sub-microsecond graph lookups and 34–58ms semantic search make it fast enough for real-time decision-making. The architecture is grounded in Cowan's three-tier working memory model—sensory buffer, working memory, long-term memory—with a hybrid decay model (exponential + power-law) from cognitive research. Shodh integrates with Claude Code, Cursor, Windsurf, Claude Desktop, VS Code (Continue extension), LangChain, LlamaIndex, OpenAI SDK, ROS2, Zenoh, and Docker. It offers a REST API, 45 MCP tools (per current docs), and client libraries for npm, PyPI, and crates.io. The tool is fully offline, open source (Apache 2.0), and backed by published research and 1,089 tests. Compared to cloud-first alternatives like mem0, Zep, or cognee—which call an LLM on every memory operation—Shodh delivers deterministic, private, and cost-free memory operations. It's designed for privacy-first, deterministic memory on edge hardware where latency and data sovereignty matter.
Behind the Verdict
Most memory tools treat memory as a vector-search problem bolted onto an LLM. Shodh Memory takes a different route: it uses Hebbian learning and decay curves to mimic biological memory, which makes it faster and cheaper than LLM-based approaches. But that architecture also means it's not for everyone—you need to be comfortable with a developer-first tool. We'd reach for Shodh when building autonomous agents that must run offline, like robotics controllers on ROS2 or edge devices on Raspberry Pi. The sub-millisecond lookups and zero LLM calls make it a fit for real-time decision loops where calling an API would add unacceptable latency or cost. If you're prototyping an agent that just needs to remember conversation context, the learning curves and graph features might be overkill—a simple KV store could do. Where Shodh really shines is in deterministic behavior. Because memory operations never touch an LLM, the same input always produces the same memory state. That's a huge plus for auditability in regulated environments. But it also means you lose the semantic flexibility that LLM-based summarization offers—Shodh won't interpret or enrich memories for you. Compared to mem0 or Zep, Shodh is the clear winner on privacy and cost. Those cloud-first tools call an LLM per operation, which adds latency and recurring fees. Shodh is free and local. However, if you need team-shared memory or a hosted management UI, Shodh's local-first design is a limitation—there's no sync or web dashboard out of the box. In practice, Shodh's MCP integration means it drops into Claude Code and Cursor smoothly. The docs are solid, and the Apache 2.0 license lets you modify it freely. But be ready for a Rust binary and CLI/API workflow—there's no GUI. If you're a developer comfortable with
Researching Shodh Memory? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Shodh Memory actually fits — and what changes day-one when you adopt it.
You want Claude Code to remember preferences and past decisions across sessions without manual memory management.
Outcome: Add the MCP server with one command, and Claude Code gains persistent, deterministic memory that improves recall with use.
Your robot needs to remember learned paths and object associations while operating fully offline.
Outcome: Deploy Shodh on the robot's onboard computer, store memory locally, and enable causal retrieval for fault-tracing during autonomous operation.
You're building a monitoring system on a Raspberry Pi with no internet connection and need semantic memory.
Outcome: Install the Python package or binary, and the Pi runs semantic search in 34-58ms, keeping all data on-device.
Use Cases
- Give Claude Code persistent memory across sessions without manual summarization
- Enable a ROS2 robot to remember learned paths and object associations offline
- Build an agent that diagnoses code issues by recalling past decisions and their causes
- Create a local-first knowledge base for a personal AI assistant that never phones home
- Run semantic memory on a Raspberry Pi for an air-gapped monitoring system
- Provide a robotics agent with causal memory for fault-tracing
Limitations
- Shodh Memory is a local-first persistent memory system that explicitly avoids using any LLM in the memory loop, making it distinct from LLM-dependent memory tools.
- It relies on Hebbian learning and decay curves, and is delivered as a binary or via package managers (npm, PyPI, crates.io) with optional Docker for server mode.
- It is designed for edge, robotics, and offline use, but is not a hosted service, so cross-machine synchronization is not provided out of the box.
- Setup requires manual installation per environment.
as of 2026-08-21
Verification history
We have re-verified Shodh Memory 7 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 7 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Shodh Memory tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0
Ideal for
Developers and small teams who want full control over their memory system and are comfortable with a CLI/API.
What this tier adds
Starting tier: free, Apache 2.0 licensed, includes all features — no paywall, no feature gating.
Enterprise
Contact
Ideal for
Organizations needing support and advanced features for production deployments of Shodh Memory.
What this tier adds
Adds dedicated support and likely advanced features; pricing is contact-based.
Where the pricing makes sense
The company stage and team size where Shodh Memory's pricing actually pencils out — and where peers do it cheaper.
Shodh Memory is free and open source (Apache 2.0), making it a cost-effective choice for developers comfortable running their own memory infrastructure. Unlike cloud memory services like mem0 or Zep that charge per memory operation or per token, Shodh has zero ongoing costs. Enterprise support is available for teams needing help, but most users can get everything from the open-source version.
Setup time & first value
How long it actually takes to get something useful out of Shodh Memory — broken out by persona, not the marketing-page minute.
For developers: add MCP server to Claude Code with one command — done in minutes. For Python: pip install, then use Python API immediately. For Docker: run the container with a volume mount, then start storing memories via REST API. For binary: download, chmod +x, run.
Switching to or from Shodh Memory
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From mem0: Export your memories via REST API and import into Shodh using its /api/remember endpoint.
Integrations
Resources & Guides
Tutorials & Learning
Official links
Tools that pair well with Shodh Memory
Common stack mates teams adopt alongside Shodh Memory, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Shodh Memory vs Spider Cloud
Choose Shodh Memory if you need deterministic, local-first memory for autonomous agents with no LLM dependency and privacy—perfect for robotics and edge. Choose Spider Cloud if you need fast, reliable web scraping and crawling to feed real-time data into AI agents or RAG pipelines. They solve entirely different problems; the right pick depends on whether your bottleneck is memory persistence or data acquisition.
Shodh Memory vs Presto Voice
These tools serve completely different markets. Shodh Memory is a developer-first, offline memory system for AI agents — ideal for privacy-conscious builders and edge robotics. Presto Voice is a drive-thru automation platform for QSR chains, focused on revenue lift. Your choice depends entirely on your domain: building autonomous agents or running a restaurant.
Shodh Memory vs Temporal Ai
For developers building autonomous agents that need deterministic, local memory without LLM overhead, Shodh Memory is a unique fit. Temporal AI is the standard for durable execution in production workflows — if your priority is reliability across crashes and retries, choose Temporal. They solve different problems: Shodh is a memory store; Temporal is an orchestration engine.
Alternatives to Shodh Memory
View allFrequently Asked Questions
Categories
Topics
Used Shodh Memory? Help shape our editorial sentiment research.