---

name: integrations-and-migration description: "Use MemMachine framework integrations and migration tooling for LangGraph, LangChain, LlamaIndex, CrewAI, AWS Strands, OpenClaw, Dify, n8n, FastGPT, example agents, and ChatGPT/export imports." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Integrations And Migration

Use this sub-skill when a task asks to connect MemMachine to an agent framework or platform, adapt an example memory-powered agent, or migrate conversation exports into MemMachine. Common signals include LangGraph, LangChain, LlamaIndex, CrewAI, AWS Strands, OpenClaw, Dify, n8n, FastGPT, simple chatbot, OpenAI/Qwen examples, ChatGPT export, OpenAI export, LoCoMo, and conversation history import.

Route Within This Sub-skill

• Read framework-integrations.md for integration patterns, prerequisites, validation checks, and route selection across supported frameworks/platforms.
• Read migration-tools.md for ChatGPT/export migration planning, local dry-runs, upload prerequisites, and duplicate/rate limit handling.
• Read troubleshooting.md for missing framework packages, server/auth errors, malformed export files, duplicate memories, and unsafe live operations.
• Run chatgpt_export_probe.py to inspect a local export JSON shape without uploading anything.

Integration Decision Flow

1. Identify the host framework/platform and whether it uses Python, TypeScript, HTTP, plugin manifests, or MCP.
2. Confirm a MemMachine API server or cloud endpoint exists and the user can provide project context and auth.
3. Pick the lightest integration path:
   • Python app code: Python SDK or framework wrapper.
   • LangGraph graph: MemMachineTools or add/search tool factories.
   • Tool-capable local clients: MCP stdio/HTTP.
   • Browser/no-code platforms: platform-specific plugin or REST integration.
   • Node/TypeScript app: TypeScript REST client.
4. Start with read-only health/config checks; run add/search only after writes are approved.

Migration Decision Flow

1. Run a local-only export probe.
2. Decide what memories should be imported and which metadata should separate user/session/source.
3. Create or confirm the target MemMachine project.
4. Upload in small batches with retry/rate-limit handling.
5. Verify with targeted searches and counts.
6. Record duplicate/skipped/failed rows without exposing private conversation content unnecessarily.

Cross-links

• Use python-sdk-and-cli for Python SDK method details, CLI command flags, filters, and LangGraph helper signatures.
• Use server-configuration-and-memory-engines for self-hosted server, storage, provider, REST, and MCP configuration.
• Use typescript-rest-client for Node/TS package usage and TypeScript app examples.

Safety Rules

• Do not run provider-backed example agents, no-code plugin actions, or live memory uploads without explicit endpoint, credentials, target project, and side-effect approval.
• Do not paste private conversation export contents into answers. Summarize counts, schema problems, and redacted examples.
• Treat benchmark/evaluation scripts as reference-only unless the user has provided datasets, services, provider credentials, and runtime budget.