Memara n8n Community Node

n8n community package n8n-nodes-memara integrates Memara’s semantic memory API. Version 2.x exposes five action nodes (one operation each) so you can pick the right step from the node panel without an operation dropdown. From 2.1.0 onward it also ships three polling triggers aligned with common Zapier-style patterns (with the API limitations described below).

Features

Action nodes

• Memara Create Memory — store content with optional title and tags
• Memara Search Memories — semantic search with limit
• Memara Update Memory — update by memory ID
• Memara Delete Memory — delete by memory ID
• Memara List Memories — paginated list (limit / offset)

Polling triggers (2.1+)

• Memara Trigger: New Memory — polls GET /api/v1/n8n/memories; first successful poll establishes a baseline (no run), then new items are detected by created_at and same-second id handling within the configured list window.
• Memara Trigger: Memory Updated — best-effort: compares a fingerprint of list fields per memory only for memories returned in the current list window; not a guarantee for the entire space.
• Memara Trigger: Search Match — polls POST /api/v1/n8n/search; emits memories whose ids newly appear in the top-N search results. Similarity scores are not returned on this API path, so threshold filtering is not supported.

Shared credentials — one Memara API credential for all nodes and triggers.

AI tools — action nodes are usable as tools where your n8n version supports it (triggers are not tools).

Polling costs and limits

Each poll is a real Memara API call and counts toward your plan limits. Triggers only see memories inside the fetched page (limit ≤ 100 for list, ≤ 50 for search). If more new memories arrive between polls than fit in that window, some may not be observed—design workflows and limits accordingly.

Installation

Option 1: Community Nodes (recommended)

1. In n8n: Settings → Community Nodes
2. Install package: n8n-nodes-memara
3. Restart n8n if prompted

Option 2: npm

npm install n8n-nodes-memara

Option 3: Local development

git clone https://github.com/memara-memory/n8n-nodes-memara.git
cd n8n-nodes-memara
npm install
npm run build
npm link

cd ~/.n8n/nodes
npm link n8n-nodes-memara

Configuration

1. API key

1. Open Memara → Settings → API Keys
2. Create a key (format mk_live_...)

2. Credential in n8n

1. Create credential Memara API
2. Paste API Key
3. Optionally override Base URL (default https://api.memara.io)
4. Test the connection

3. Add nodes to a workflow

1. Add node and search for Memara
2. Choose an action node (Memara Create Memory, Memara Search Memories, etc.) or a trigger (Memara Trigger: New Memory, etc.)
3. Attach the same Memara API credential to each node

Migrating from v1.x

v1 shipped a single node with internal type memara and parameters resource, operation, and operation-specific fields.

v2 removes that node. Workflows break until you replace each old node with the matching v2 node:

Triggers (2.1+):

Remove resource and operation from parameters; keep only the fields that node needs (same names as before: content, title, tags, memoryId, query, limit, offset).

Search / List output: Each run still produces one item whose json is the array of memories (same shape as v1’s search/list output for typical workflows).

Usage examples (parameters)

Create (Memara Create Memory)

{
  "content": "Customer reported billing issue #12345. Resolved with 10% credit.",
  "title": "Customer issue resolution",
  "tags": "customer-service, billing"
}

Search (Memara Search Memories)

{
  "query": "billing problems with customers",
  "limit": 5
}

Update (Memara Update Memory)

{
  "memoryId": "mem_...",
  "content": "Updated content",
  "title": "Optional title",
  "tags": "tag1, tag2"
}

Delete (Memara Delete Memory)

{
  "memoryId": "mem_..."
}

List (Memara List Memories)

{
  "limit": 50,
  "offset": 0
}

Common workflow patterns

1. Support KB: Webhook → Memara Search Memories → branch on results → Memara Create Memory for new resolutions.
2. Content assist: Schedule or external trigger → Memara Search Memories → Code / LLM → Memara Create Memory.
3. React to new memories: Memara Trigger: New Memory → notify Slack / CRM / internal API (each item is one memory payload plus _memaraTrigger).

Example workflow JSON: workflows/customer-support-example.json. Polling trigger starter: workflows/new-memory-trigger-example.json.

Parameters reference

Troubleshooting

• Invalid API key: Confirm key in Memara dashboard; default API URL is https://api.memara.io.
• Node not listed: Reinstall the package, restart n8n, check npm list n8n-nodes-memara.

Docs: memara.io/docs/integrations/n8n · Support: support@memara.io

Development

Local (no Docker)

git clone https://github.com/memara-memory/n8n-nodes-memara.git
cd n8n-nodes-memara
npm install
npm run build
npm run verify:pack
npm run lint

Docker (recommended parity with CI / publish)

Uses docker-compose.n8n.yml and scripts/docker-n8n.sh. From packages/memara-n8n-node:

After a local build, npm run verify:pack runs npm pack and checks the archive for unexpected paths (for example stray third-party dist artifacts). The same step runs inside docker-n8n.sh ci before lint.

n8n community verification scanner (queries the public npm package n8n-nodes-memara—useful after you publish a version):

npm run docker:scan-npm
# or locally: npm run scan:npm

Team: mirror to GitHub from the monorepo (same idea as Memara Python SDK)

Maintainers can sync this directory to memara-memory/n8n-nodes-memara using Docker: git runs inside memara-n8n-dev in an ephemeral /tmp clone, populated with tar from /app (the bind-mounted package). You do not need a nested .git under packages/memara-n8n-node, and you do not need the memara monorepo git root to equal this folder.

1. Put a fine-grained PAT (Contents read/write on that repo; for releases also Releases write when using publish-release) in packages/memara-n8n-node/.env.github (single line).
2. From packages/memara-n8n-node:

./scripts/docker-n8n.sh github-verify-pat
./scripts/docker-n8n.sh github-sync
# non-interactive:
./scripts/docker-n8n.sh github-sync --yes
# after sync, push annotated tag v<version> from package.json:
./scripts/docker-n8n.sh github-tag --yes

github-sync excludes build secrets and artifacts (.env*, dist, node_modules, etc.); the public repo receives sources so GitHub Actions or local builds can produce dist.

One-shot: CI, GitHub mirror, npm, and GitHub Release

From packages/memara-n8n-node (including inside the memara monorepo): bump and commit package.json version, then:

1. packages/memara-n8n-node/.env.github — GitHub PAT as above (needs Releases permission for the API step in publish-release).
2. packages/memara-n8n-node/.env.npm — npm token (same as publish below).

./scripts/docker-n8n.sh publish-release
# or: npm run docker:publish-release
# or: ./scripts/release-publish.sh --yes
./scripts/docker-n8n.sh publish-release --yes

This runs Docker CI (build → verify:pack → lint), github-sync (mirror to public main), github-tag, publish to npm, then creates a GitHub Release on the host via the REST API (PAT never printed).

If this package lives inside a parent git repo, uncommitted changes under packages/memara-n8n-node must be clean before publish-release runs (paths are scoped to that directory).

Publish to npm via Docker

Publishing is manual (nothing is published from CI in this repo unless you run it). Bump package.json version, then:

1. Put your npm access token in packages/memara-n8n-node/.env.npm (single line, no variable name). The file is gitignored.
2. From packages/memara-n8n-node:

./scripts/docker-n8n.sh publish
# or: npm run docker:publish
# non-interactive:
./scripts/docker-n8n.sh publish --yes

The script starts memara-n8n-dev if needed, runs npm run build in the container, then publishes using a temporary npm userconfig so your token is not written to the repo’s package.json or a persistent .npmrc.

GitHub Actions and npm provenance (n8n verified nodes)

For n8n Cloud verified community nodes, npm expects provenance from GitHub Actions (see Submit your node for verification). This package includes .github/workflows/publish-npm.yml, which publishes with npm publish --access public --provenance when you publish a GitHub Release (or run the workflow manually).

1. In the npm package settings, connect a Trusted Publisher to memara-memory/n8n-nodes-memara (or your fork) and this workflow, so OIDC can authenticate publishes where possible.
2. Add repository secret NPM_TOKEN with a granular Automation token that can publish n8n-nodes-memara (required until Trusted Publisher fully replaces token auth for your account).
3. Keep GitHub package.json version and git tags aligned with each npm release so reviewers can verify source.

If publish fails with 403 / two-factor authentication: npm requires either an interactive publish with OTP, or a granular access token that includes Publish for n8n-nodes-memara and the “Bypass two-factor authentication” option (or equivalent automation permission). Classic tokens alone are often blocked when 2FA is enabled on the account.

License

MIT — see LICENSE.md.

Made with care by the Memara team.