Agents Overview
Give your AI agents a shared home for their work. MindWiki provides cloud storage, a Markdown file system, and reusable memory that compatible agents can search, read, and build on, while you access the same vault through MindWiki Cloud, the iPhone app, and the Mac app.
Research saved with one agent can become the brief another agent reads before writing or coding. Decisions survive the conversation. Sources remain beside the conclusions they support. You can inspect and edit the work yourself instead of relying on an agent's account of what it remembers.
What a connected agent can help you do
| Capability | What it makes possible |
|---|---|
| Search and read | Retrieve your prior research, decisions, preferences, and project context. |
| Capture | Save a useful conversation outcome directly to your inbox. |
| Create and update pages | Develop briefs, plans, source summaries, and handoffs in your vault. |
| Retrieve task context | Find saved pages that match a specific goal and constraints. |
| Follow relationships | Explore links, related pages, and graph data. |
| Use advanced workflows | Work with Pixel, rich documents, attachments, and publications when authorized. |
| Inspect activity | Review recent recorded tool use and sessions. |
Capabilities depend on the connection's scope, your MindWiki plan, and your vault permissions. You can start with retrieval, then enable inbox capture or broader writes as needed.
One vault for multiple agents
Connect each platform to the same MindWiki account. Each gets its own connection to the same underlying knowledge. You keep your existing pages and capture address when adding another agent.
A practical sequence is:
- Ask a research agent to save findings and source references.
- Ask a planning agent to read those pages and create a brief.
- Ask a coding agent to retrieve the brief and constraints before implementing.
- Save the implementation outcome, checks, and remaining work for the next session.
- Open the resulting pages yourself in MindWiki to review or refine them.
This is a shared workspace built from saved files and explicit retrieval. MindWiki does not automatically import every agent's chat history or make agents coordinate continuously in the background. Give each task a clear brief and save the output that the next task needs.
Share your vault conventions
Use the Agent Operating Guide to record where project decisions go, which templates to use, and which folders to leave alone. In a connected client that offers the guide tool, try “Use my MindWiki vault conventions when you save this decision.” The assistant can read the relevant section for that task. Your preferences are available to authorized clients connected to the same vault.
Choose your agent
Each link opens a dedicated guide with setup, authorization, a first useful request, verification, and troubleshooting. Choose the interface you actually use; browser connectors and local coding tools have different installation paths.
- ChatGPT — bring vault context into questions, research, and drafting.
- Claude.ai — use your saved knowledge in browser conversations.
- Grok / Grok Bot — connect Grok chat or a Bot using the setup for that client.
- Claude Desktop — connect the desktop client to your cloud vault.
- Claude Code — retrieve project context and preserve coding handoffs.
- Codex — work from saved briefs, requirements, and engineering decisions.
- Cursor — bring project knowledge into the coding workspace.
- GitHub Copilot — connect supported agent workflows to your vault.
- Gemini CLI — retrieve and save knowledge from the terminal.
- Cline — use shared context during development tasks.
- Windsurf / Cascade — connect coding work with saved project knowledge.
- Devin — provide persistent context and capture useful task outcomes.
- Muse — connect through the connector controls offered by your account.
- Muse Code — use your vault in supported coding workflows.
- Other MCP Clients — configure compatible clients beyond these guides.
Configuration-capable coding agents have a copyable setup request in their guide. Paste it into the agent and let it configure the connection where supported. You still complete MindWiki authorization yourself. Browser-based agents generally require you to add the connector through their own controls.
Connect and verify
The shared remote endpoint is:
https://api.mindwiki.io/mcpFollow your client guide, authorize the connection in MindWiki, and return to the agent. Existing users should sign in to their existing account. New users following the agent authorization flow choose their MindWiki email name and save their generated sign-in access key privately.
Then ask for a real read:
Search my MindWiki for a project I am working on. Read a relevant page and summarize its goal and next steps. Tell me the page title and path you used.For a connection with capture access, save a short result and open it yourself in capture/. A successful read proves retrieval; a successful save proves the authorized writing path. A URL appearing in settings alone does not establish either.
Build useful memory deliberately
Save conclusions, evidence, decisions, constraints, and next actions. Ask the next agent to retrieve them at the beginning of its task. Link frequently used sources from a project brief so any authorized client can retrieve the same starting point.
Use Mission Control on Mac to review activity and proposals. Manage individual grants in Connections. Your sign-in access key belongs on MindWiki's own sign-in screens; MCP clients use their connection credentials.
Free includes one external AI connection with core vault tools. To see what your connection can use, review MCP tool access and your agent's setup guide.
Next: Shared Agent Memory, Project Workflows, and Integrations.