# openclaw-supermemory: Long-Term Memory and Profile Persistence for the OpenClaw Agent > openclaw-supermemory is a plugin for the OpenClaw agent that automatically captures every conversation turn to a cloud memory service and injects relevant past context before each AI response. It requires a Supermemory Pro subscription and OpenClaw version 2026.5.7 or later. **supermemoryai/openclaw-supermemory** — OpenClaw Supermemory lets to have long-term memory and recall for your openclaw agent. - Repository: https://github.com/supermemoryai/openclaw-supermemory - Website: https://supermemory.ai/docs/integrations/clawdbot - Stars: 796 · Forks: 93 - Language: TypeScript - License: not declared - Published: 2026-09-14 · Updated: 2026-09-14 · Language: en - Canonical page: https://hysenlabs.com/projects/supermemoryai-openclaw-supermemory ## What the plugin adds to OpenClaw and who it is for OpenClaw is an AI coding agent. Without a memory plugin, each session starts with a blank context. The agent has no knowledge of previous conversations, past decisions about a codebase, or accumulated preferences about how the user wants to work. openclaw-supermemory solves this by intercepting every AI turn in two directions. Before the turn, it queries the Supermemory API for memories relevant to the current conversation and injects them into the context. After the turn, it sends the conversation to Supermemory for storage. The README describes the result: the AI sees the user's profile and semantically similar past conversations at the start of every turn. The target user is an OpenClaw power user who works on ongoing projects and finds it frustrating to re-explain context at the start of each session. The plugin is not a generic memory layer for any AI tool: it is specifically designed for OpenClaw and requires OpenClaw version 2026.5.7 or later as a peer dependency. ## Installation and initial setup Installing the plugin requires a running OpenClaw environment with gateway access. The install command is: ```bash openclaw plugins install @supermemory/openclaw-supermemory ``` After installation, run the setup command to configure the API key: ```bash openclaw supermemory setup openclaw gateway restart ``` The setup command prompts for a Supermemory API key, which you get from app.supermemory.ai. The gateway restart applies the configuration. For users who want to configure all available options interactively: ```bash openclaw supermemory setup-advanced openclaw gateway restart ``` The advanced setup exposes container tag, auto-recall, auto-capture, capture mode, and custom container tag settings. The API key can also be set through an environment variable instead of the interactive setup: ```bash export SUPERMEMORY_OPENCLAW_API_KEY="sm_..." ``` The README notes that the plugin requires Supermemory Pro or above. The README does not document what happens if the API key is from a free or lower-tier account. ## Auto-recall, auto-capture, and the memory retrieval mechanism Once installed, the plugin operates without user intervention in normal use. Auto-recall and auto-capture are both enabled by default. Auto-recall fires before every AI turn. The plugin queries Supermemory with the current conversation context and receives the most relevant stored memories, up to the maxRecallResults limit (default 10). These memories are injected into the AI's context so the model sees them before generating a response. Auto-capture fires after every AI turn. The full conversation turn is sent to Supermemory, which extracts relevant information, deduplicates it against existing memories, and stores it. The README describes the extraction and deduplication as running in the Supermemory cloud, not locally. The profileFrequency setting controls how often the full user profile is injected rather than just semantically relevant memories. The default is every 50 turns. The captureMode setting has two values: "all" filters out very short texts, while "everything" captures all turns without filtering. The README explains that everything runs in the cloud. No local vector database or embedding model is required on the OpenClaw host. ## Custom container tags and memory routing By default, all memories for a given OpenClaw instance are stored in a single container tagged openclaw_{hostname}. The custom container tag feature allows splitting memories into separate named containers such as work, personal, or bookmarks. Enabling custom containers requires two configuration changes: setting enableCustomContainerTags to true and defining the containers array with entries that each have a tag and a description. The customContainerInstructions field provides the AI with guidance on which container to use for a given piece of information. Once configured, the AI automatically picks the right container based on the provided instructions when saving a memory. All four AI tools (supermemory-save, supermemory-search, supermemory-forget, and supermemory-profile) support a containerTag parameter for explicit routing. The README also documents an older naming convention for the tools: supermemory_store, supermemory_search, supermemory_forget, and supermemory_profile. These deprecated names still work in version 2.1.18 but are scheduled for removal in version 3.0. Tool allowlists that reference the old names with underscores should be updated to the hyphenated names. ## Slash commands and CLI for manual memory operations The plugin adds three slash commands that users can invoke directly in an OpenClaw session: - /remember followed by text immediately saves that text to memory without any AI processing. - /recall followed by a query searches memories by similarity and returns results with similarity scores. - /supermemory-index is an agent skill that explores the current project's codebase, extracts focused memories about it, and saves them. The README notes this differs from /remember because it actively crawls the codebase rather than saving a single piece of text. For operations outside the agent session, the CLI provides equivalent commands: ```bash openclaw supermemory status openclaw supermemory search openclaw supermemory profile openclaw supermemory wipe ``` The wipe command deletes all stored memories and requires confirmation. The status command shows the current configuration without modifying anything. These CLI commands are useful for auditing what the plugin has stored and for clearing memory before switching to a new project or user context. ## Configuration reference and deployment considerations The full configuration lives in ~/.openclaw/openclaw.json under plugins.entries.openclaw-supermemory.config. The README provides a comprehensive table of configuration options. Key settings beyond the API key include: - baseUrl: defaults to https://api.supermemory.ai. Set this to a self-hosted Supermemory instance URL if you are running your own deployment. - containerTag: the root memory namespace, defaults to openclaw_{hostname}. Changing this after memories are stored does not migrate existing memories. - maxRecallResults: defaults to 10. Higher values inject more context but consume more tokens. - autoRecall and autoCapture: both default to true. Setting either to false disables the corresponding automatic behaviour. A real alternative within the OpenClaw ecosystem is openclaw-mem0, listed in the related searches for this project. The difference is the memory backend: this plugin uses Supermemory cloud, while openclaw-mem0 uses Mem0. Both provide automatic recall and capture, but they differ in pricing model and in what the underlying service stores. Supermemory explicitly builds a persistent user profile; the README for this plugin describes that profile as a set of persistent facts plus recent context. The last push to the repository was on 2026-09-09. ## Conclusion openclaw-supermemory suits OpenClaw users who have a Supermemory Pro subscription and want their agent to accumulate a persistent user profile and recall past conversations automatically. It is not for users who want memory that runs entirely on local infrastructure: the README makes clear that Supermemory handles extraction, deduplication, and profile building in the cloud, with no local model option described. The self-hosted base URL option in the configuration allows pointing at a self-hosted Supermemory instance, but operating such an instance is outside this plugin's scope. The last push was on 2026-09-09. The package.json declares an MIT licence. ## FAQ ### How does Supermemory work in the openclaw-supermemory plugin? The plugin queries Supermemory before every AI turn and injects the most relevant stored memories as context. After every turn, it sends the conversation to Supermemory for extraction and storage. All extraction, deduplication, and profile building happens in the Supermemory cloud. ### What is OpenClaw's memory and how does it work through this plugin? openclaw-supermemory gives OpenClaw a persistent user profile and semantic search over past conversations. Auto-recall injects relevant past context before each AI turn, and auto-capture stores new conversations after each turn. The memory is namespaced by container tag and runs entirely in the Supermemory cloud. ### Is Supermemory free to use with this plugin? The README states that openclaw-supermemory requires Supermemory Pro or above. A free Supermemory tier is not described as sufficient to use this plugin. ## Sources - [Issues](https://github.com/supermemoryai/openclaw-supermemory/issues) - [Project website](https://supermemory.ai/docs/integrations/clawdbot) - [README](https://github.com/supermemoryai/openclaw-supermemory/blob/main/README.md) - [supermemoryai/openclaw-supermemory on GitHub](https://github.com/supermemoryai/openclaw-supermemory) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/supermemoryai-openclaw-supermemory