Cloudflare Workers AI runs 50+ open-source AI models on serverless GPUs, billed per use in Neurons from Workers or the REST API.
Freemium
agent-memory
A local long-term memory runtime for AI agents that stores memories as Markdown files, ranks recall with a rebuildable index and needs no API key.
What is agent-memory?
agent-memory, from tigerless-labs, gives AI agents a memory that survives after a session closes. Each memory is one Markdown file in a shared store, and a SQLite index beside it works only as a cache that can be deleted and rebuilt with mem rebuild. Any agent able to execute shell commands, including Claude Code, Codex CLI and Muse Code, can read and write the same store. Recall is local and ranked using BM25 over an FTS5 index, with an optional vector plugin, and it answers with file paths so the agent opens each hit only as deep as it needs: abstract, outline or full file. Writes fire at conversation boundaries instead of waiting for the agent to remember, and a sleep-time pass consolidates memories, with deletions arriving only as proposals you confirm. There is no LLM client inside the library, so it has no API key and no billing of its own. It is open source and free to run.
Use cases
- Share memory between Claude Code and Codex
- Keep agent notes as readable Markdown
- Search past agent knowledge locally
- Consolidate old agent memories automatically
- Answer what was known as of a date
- Move agent memory between tools
Who it's for
Why use agent-memory
- Memory stays readable and portable because it is plain Markdown you can browse with ls and grep or keep in git.
- No API key or billing is involved: the library borrows its judgement from the host agent's CLI, and retrieval happens locally.
- Different agents share one store, so switching from Claude Code to Codex does not lose notes.
- Writes do not depend on the agent remembering to save, because they trigger at conversation boundaries and the full trace is copied first.
agent-memory features
- One Markdown file per memory with frontmatter and links
- Disposable SQLite index rebuilt with mem rebuild
- BM25 recall over FTS5 with optional vector fusion
- Recall that returns paths at abstract, outline or full depth
- Automatic distillation at conversation boundaries
- Sleep-time consolidation with confirmed deletions
- Supersede history and recall --as-of queries
- mem CLI plus a stdio MCP server through mem-mcp
- Hooks for SessionStart, Stop, SessionEnd and PreCompact
agent-memory integrations
How to get started with agent-memory
- Install the runtime and run mem setup so your host agent, such as Claude Code or Codex CLI, gets the hooks and skill.
- Work normally; at session boundaries the runtime distils the conversation into Markdown memories in the store.
- At the next session start, MEMORY.md is injected, and the agent can run mem recall to get a ranked list of paths.
- The agent reads only what it needs with mem read at the outline, abstract or full level.
- A sleep-time pass consolidates memories and proposes deletions that you confirm.
agent-memory limitations
- Local only: there is no hosted service
- Context, sleep and the proposal ledger work only through the CLI; MCP does not expose them yet
- Automatic sleep-time runs happen only after the next session event unless you install a cron entry
agent-memory pricing
agent-memory is open source, so you can use and self-host it for free.
Prices can change and may vary by country or exclude tax. Check the official pricing page.
agent-memory alternatives
The main agent-memory alternatives listed on Debuly are Mem0, Letta, Zep, Cognee, Supermemory and ReMe. The table compares pricing and platforms; order is A–Z.
| Tool | Pricing | Starting price | Platforms |
|---|---|---|---|
| agent-memory | Open source | — | CLI |
| Cognee | Freemium | $1, usage-based | Web, API, Self-hosted, MCP server |
| Letta | Freemium | $20/mo | CLI, Web, API |
| Mem0 | Freemium | $19/mo | API, Web, Self-hosted, CLI |
| ReMe | Open source | — | CLI, Self-hosted, Web, API |
| Supermemory | Freemium | $19/mo | API, Web, Self-hosted |
| Zep | Freemium | — | API |
Cognee
Open-source memory engine for AI agents that combines vector search, knowledge graphs and relational storage, with a hosted cloud.
FreemiumVisit Cognee (official site)
Letta
Platform, formerly MemGPT, for building stateful AI agents with persistent memory, from a terminal app, SDK or API.
FreemiumVisit Letta (official site)
Mem0
Memory layer for AI agents and apps, available as an open-source package or a managed platform that keeps user context across sessions.
FreemiumVisit Mem0 (official site)
ReMe
Open-source memory layer for AI agents that stores knowledge as local Markdown files with hybrid search and a linked graph.
Open sourceVisit ReMe (official site)
Supermemory
Memory and context API for AI agents with user profiles, a memory graph, retrieval and connectors, billed by usage.
Zep
Managed memory and context platform for AI agents, built on the Graphiti knowledge graph, with Python, TypeScript and Go SDKs.
FreemiumVisit Zep (official site)
agent-memory FAQ
Is agent-memory free?
Yes. It is open source, runs locally and has no API key or billing of its own. No paid plans are listed.
Which agents can use agent-memory?
The README names Claude Code, Codex CLI and Muse Code, and says anything that can run a shell command can share the same store.
Does agent-memory have an API or MCP server?
It has the mem command-line tool and a stdio MCP server, mem-mcp, which exposes calls such as memory_recall, memory_read and memory_record. Setup adds it when you pass --mcp.
What happens if I delete the SQLite index?
Nothing is lost. The index is a cache, and running mem rebuild regenerates it from the Markdown files, which are the source of truth.
Does agent-memory send my data to a model provider?
Judgement is borrowed from the host agent's CLI by default. An endpoint such as Vertex AI or an OpenAI-compatible URL is opt-in, and the library carries no credentials of its own.
What are alternatives to agent-memory?
Alternatives include Mem0, Letta, Zep and Cognee, which also give agents persistent memory.
Sources & verification
Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.
| Source | Type | Accessed |
|---|---|---|
| github.com/tigerless-labs/agent-memory | official |
Image credits
- Logo: from avatars.githubusercontent.com · fetched · Vendor trademark, used only to identify the product; removed on request
- Cover: none collected yet, so a generated tile is shown.
Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.
More in Developer platforms 4
All developer platforms toolsGPU cloud for AI with on-demand NVIDIA instances and reserved 1-Click Clusters for training, fine-tuning and inference.
PaidDecentralized cloud compute on Kubernetes for deploying apps, LLM interfaces and AI workloads, paid on-chain with the LNQ token.
PaidControl plane for production AI: monitor traffic, fine-tune models and version workflows and agents from live usage.
Paid
