ClawHub

mnemospark

mnemospark is an OpenClaw plugin that gives agentic systems instant, secure access to cloud storage, compute, and proprietary datasets paid via x402 with USDC on Base. Wallet and go.

Audits

Warn
ClawScanWarn

Agentic behavior and permission review.

Static analysisReview

Pattern checks against bundled files.

VirusTotalstale

Multi-engine malware detections and file reputation.

Install

openclaw plugins install clawhub:mnemospark

mnemospark

Wallet and go. 💙
No forms. No email. Just Base. 💙

mnemospark is an agentic service layer for OpenClaw plus a standalone x402 payment + verification backend. It enables encrypted, authenticated, fully autonomous cloud and data workflows—paid via x402 with USDC on Base. No human onboarding. No API keys. The blockchain transaction is the record.

For the story behind the product—agent-native encrypted cloud storage on x402 with USDC on Base—see the launch post on Paragraph.

What is mnemospark?

mnemospark connects OpenClaw agents to cloud workflows with wallet-native auth + payment rails.

  • Wallet-native authentication for storage/API actions
  • x402-native payments with USDC on Base
  • Agent-first flow (quote → pay → provision)
  • Encrypted payload support and structured request signing
  • Built for automation (cron-friendly, idempotent-friendly workflows)

Quick Start

1) Install the plugin in OpenClaw

openclaw plugins install mnemospark --dangerously-force-unsafe-install
openclaw gateway start

Plugin registration is done by openclaw plugins install mnemospark. The install also bundles the skills/mnemospark skill package so the main agent can delegate mnemospark workflows. Tell your agent to read it!

By default the plugin uses the mnemospark API at https://api.mnemospark.ai.

2) Restart gateway after updates

openclaw gateway restart

3) Use slash commands in OpenClaw chat

  • /mnemospark wallet → wallet status; /mnemospark wallet help → commands; /mnemospark wallet create → new key (backs up existing default wallet.key if present)
  • /mnemospark cloud help → storage command guide

OpenClaw Install Warning

If OpenClaw shows a warning about dangerous code patterns when installing or updating mnemospark—often mentioning shell execution (child_process), environment variables, and network access—here is what is going on.

mnemospark is an OpenClaw plugin that talks to your configured mnemospark backend, runs a local HTTP proxy for storage workflows, and can invoke the openclaw CLI and system tools when needed (for example creating archives with tar or running npm when you use the update command). Those features use the same low-level Node.js APIs—child_process and fetch—that security tools also associate with risky software, so the installer may warn you even when the behavior is intentional and benign.

We also read environment variables you set on purpose (such as MNEMOSPARK_BACKEND_API_BASE_URL, MNEMOSPARK_PROXY_PORT, or wallet-related settings) so you can configure the plugin without editing code. Automated scans sometimes flag “environment access + network” as a possible credential-stealing pattern. In mnemospark, that combination exists because the plugin is configurable and networked by design, not because we are harvesting your unrelated secrets.

mnemospark is open source. If you want extra assurance, review the repository, search for child_process, process.env, and fetch, and run your own tests in a safe environment. The warning helps keep the ecosystem safe; for mnemospark it reflects capabilities, not a finding of malicious intent.

Have ChatGPT or your favorite LLM evaluate the mnemospark codebase. Here's a starter link.

Slash command reference

Syntax

  • Chat / OpenClaw (primary): use key:value for arguments (for example wallet-address:0x…, region:us-east-1). key=value is also accepted.
  • Agent-driven workflows and CLI-style usage: prefer --parameter value (for example npx mnemospark or shell scripts).

Top-level routes (after /mnemospark)

  • /mnemospark or /mnemospark help — overview
  • /mnemospark cloud … — cloud storage (see below)
  • /mnemospark wallet … — wallet helpers

/mnemospark cloud — includes: cloud / cloud help, backup, price-storage, upload, payment-settle, ls, download, delete, op-status. Optional flags include async:true, orchestrator:inline|subagent, timeout-seconds:<n>. For the full cloud guide, run /mnemospark cloud help in chat.

/mnemospark wallet

  • /mnemospark wallet — address, balance, and key file path
  • /mnemospark wallet help — command list and funding link
  • /mnemospark wallet create — create a new wallet (and backup an existing default key file)
  • /mnemospark wallet export — export private key for backup (sensitive)

Core Commands

Use via /mnemospark cloud … (or /mnemospark wallet …) in OpenClaw chat. Prefer key:value in chat; use --wallet-address style for scripted or agent-driven flows.

Get a storage quote

/mnemospark cloud price-storage wallet-address:<addr> object-id:<id> object-id-hash:<sha256> gb:<gb> provider:aws region:us-east-1

Use other regions by changing provider: and region: (defaults: aws / us-east-1).

Upload using quote

/mnemospark cloud upload quote-id:<quote-id> wallet-address:<addr> object-id:<id> object-id-hash:<sha256>

List objects

/mnemospark cloud ls wallet-address:<addr> object-key:<object-key>

Download object

/mnemospark cloud download wallet-address:<addr> object-key:<object-key>

Delete object

/mnemospark cloud delete wallet-address:<addr> object-key:<object-key>

Payment Model (x402 + Base)

mnemospark follows a quote-and-pay execution model:

  1. Agent requests a quote.
  2. Agent provides wallet-native + payment authorization.
  3. Backend verifies payment/auth context.
  4. Storage action executes.

The blockchain transaction is the payment record.

Security Model

  • Wallet-signed request authentication
  • Encrypted payload flow (AES-256-GCM request contract)
  • Idempotency-aware upload/payment handling
  • Structured auth/payment event logging for traceability

Keep wallet private keys secure. Anyone with the key can control wallet-authorized actions.

Environment variables

Optional unless noted. All names use the MNEMOSPARK_ prefix.

VariablePurpose
MNEMOSPARK_BACKEND_API_BASE_URLBase URL for the mnemospark backend API (required for the local HTTP proxy to forward storage calls). Default: https://api.mnemospark.ai (production). Set this variable only to override—for example staging or a custom API Gateway URL.
MNEMOSPARK_PROXY_PORTTCP port for the mnemospark HTTP proxy (default 7120).
MNEMOSPARK_DOWNLOAD_DIRDirectory where the proxy writes downloaded objects (default ~/.openclaw/mnemospark/downloads/).
MNEMOSPARK_WALLET_KEYPath to the wallet private key file when not using the default ~/.openclaw/mnemospark/wallet/wallet.key.
MNEMOSPARK_REMOVE_BACKUP_FILEAfter a successful cloud upload, delete the local backup archive under ~/.openclaw/mnemospark/backup/. Default when unset: remove the file. Set to 0, false, no, or n to keep it; 1, true, yes, or y to remove.
MNEMOSPARK_DISABLEDSet to true or 1 to disable plugin registration.
MNEMOSPARK_DISABLE_SQLITESet to 1 to disable local SQLite (state.db); cloud commands that need local state will fail.
MNEMOSPARK_SQLITE_STRICTSet to 1 so certain SQLite consistency checks (e.g. friendly-name verification after upload) throw instead of warning.
MNEMOSPARK_PROXY_VERBOSE_404When 1, true, or yes, the local HTTP proxy includes a message field on 404 responses describing supported paths. Default (unset) is a generic JSON body { "error": "Not found" } only (reduces reconnaissance).
MNEMOSPARK_CRON_AGENT_IDOpenClaw agent id used for the monthly renewal cron (default mnemospark-renewal).
MNEMOSPARK_AGENT_IDOpenClaw agent id for interactive wallet/cloud CLI via the dedicated-agent runbook (default mnemospark). Distinct from MNEMOSPARK_CRON_AGENT_ID.
MNEMOSPARK_CRON_NODE_BINAbsolute path to node for dedicated-agent exec allowlists and renewal cron payloads (default /usr/bin/node).
MNEMOSPARK_DISABLE_OPENCLAW_PREREQSet to 1 to skip automatic runbook application everywhere it runs (plugin load, CLI install/update; for advanced debugging only).

Typical Workflow

  1. Install plugin
  2. Fund Base wallet with USDC
  3. Request quote
  4. Execute upload (supports up to 5 GB files - multipart uploads coming soon!)
  5. Confirm/list/download/delete as needed

mnemospark Exec Approvals Runbook

On OpenClaw 2026.4.x, mnemospark applies two dedicated-agent runbooks when you install or update (including npx mnemospark install, npx mnemospark update, openclaw plugins install, and when the gateway loads the plugin):

  1. mnemospark-renewal (default; override MNEMOSPARK_CRON_AGENT_ID) — monthly storage renewal cron: tools.deny: ["subagents"], tools.exec.ask: "off", and /usr/bin/node allowlisted for that agent in ~/.openclaw/exec-approvals.json.
  2. mnemospark (default; override MNEMOSPARK_AGENT_ID) — interactive wallet and cloud CLI work: same tool policy and its own /usr/bin/node allowlist entry under that agent id.

The plugin merges both into agents.list via openclaw config set + openclaw config validate. OpenClaw restarts the gateway when a plugin is installed or updated; mnemospark does not call openclaw gateway restart itself. After a successful upload, mnemospark registers the monthly renewal cron only (--no-deliver, --agent, and a Command: /usr/bin/node …/dist/cli.js cloud payment-settle --renewal … message) against the renewal agent id. For interactive automation, route through the general agent (see bundled skill skills/mnemospark/references/openclaw-routing.md). Override the node path with MNEMOSPARK_CRON_NODE_BIN if your system differs.

Troubleshooting

  • Missing wallet/auth errors: verify wallet key is present and request signature headers are generated.
  • 402 payment required: expected in challenge flow; ensure client retries with payment authorization.
  • Upload/storage backend errors: verify cloud permissions (e.g. bucket access + IAM role rights).
  • Command not recognized: confirm plugin installed and gateway restarted.
  • Renewal cron / exec failures after upload: ensure mnemospark was installed or updated through the normal path so both dedicated-agent runbooks ran (gateway load, npx mnemospark install, or openclaw plugins install). If the gateway never loaded the plugin, run install again or restart the gateway.
  • Interactive Mnemospark exec denied on main: use a dedicated mnemospark agent and route with openclaw agent --agent mnemospark … (see skill reference openclaw-routing.md); renewal-only setup does not cover general plugin usage.
  • One-step operation correlation: run ./skills/mnemospark/scripts/debug-operation.sh <operation-id> (or omit ID to use latest).

Repos

  • Plugin: pawlsclick/mnemospark
  • Backend: pawlsclick/mnemospark-backend