API access to your Recall knowledge base
under review
P
Paul Richards
The API would allow you to access all your Recall data as well as add more content and edit existing content.
Log In
Sankari Nair
Merged in a post:
Recall & Openclaw
H
Heather M
I'm curious if there's a list of use cases / skills generated to use Recall & Openclaw ... I suspect there's great potential.
Sankari Nair
Merging this one in the API request as we would need an API to set this up!
Sankari Nair
Merged in a post:
Official MCP Server for Bidirectional AI Agent Integration (Claude Code, Cursor, Windsurf)
l
lars.viding007
### Description
With the rapid adoption of AI coding agents (Claude Code, Cursor, Windsurf, Codex), there's a growing need for Recall to provide a
Model Context Protocol (MCP) server
that enables bidirectional integration between Recall's knowledge base and AI agent workflows.Why MCP specifically (not just a REST API):
MCP is becoming the standard protocol for connecting AI agents to external data sources. Claude Desktop, Claude Code, Cursor, Windsurf, and other tools all support MCP natively. An MCP server would give Recall instant compatibility with the entire ecosystem — no custom integrations needed per tool.
Proposed MCP tools:
| Tool | Description |
|------|-------------|
|
recall_search
| Semantic search across the user's knowledge base with filters (tags, source type, date range) ||
recall_get_card
| Retrieve full card content by ID ||
recall_create_card
| Create a new card from agent-generated content (research results, summaries, decisions) ||
recall_tag
| Add/remove tags on existing cards ||
recall_list_recent
| List recently added or modified cards |Use cases this enables:
- Knowledge-grounded agents— An agent researching a topic checks the user's Recall KB first, giving priority to curated sources over raw web results
- Automated knowledge capture— After a research session, the agent saves key findings back to Recall with appropriate tags
- Cross-session continuity— Different agent instances (terminal, IDE, desktop) all share the same knowledge base
- Bulk organization— Ask an agent to re-tag, merge, or reorganize cards based on semantic relationships
Implementation reference:
A simple remote MCP server config would look like:
```json
{
"mcpServers": {
"recall": {
"url": "https://api.getrecall.ai/mcp",
"headers": {
"Authorization": "Bearer <RECALL_API_KEY>"
}
}
}
}
```
This pattern is already established by tools like Todoist, Asana, and Linear who ship official MCP servers. Recall's knowledge ingestion and AI summarization is best-in-class — adding MCP access would make it the central knowledge hub for AI-powered workflows rather than just a standalone app.
Competitive context:
Users are already discussing migrating to Obsidian + CLI + MCP setups because they offer programmatic access. An official MCP server would eliminate this migration pressure and position Recall as the knowledge layer for agentic AI workflows.---
l
lars.viding007
+1 from an Azure architect using Claude Code agents daily.
My workflow: I use Recall to ingest YouTube videos, articles, and technical docs into a curated knowledge base. It's fantastic for that. But my actual
work
happens through Claude Code agents running in terminals, VS Code, and Claude Desktop — each with different contexts and tasks.Right now there's a gap: my agents can search the web, read local files, and query APIs — but they can't tap into the curated knowledge I've spent time building in Recall. That means I end up manually copy-pasting context from Recall into prompts, which defeats the purpose of having AI agents.
What I need specifically:
- Read access via MCP or REST API— so my Claude Code agents can query my Recall KB as a first-pass knowledge source before falling back to web search
- Write access— so agents can push research results, summaries, and architectural decisions back into Recall for future reference
- Semantic search endpoint— not just keyword matching, but the same smart retrieval Recall's built-in chat uses
The competitive pressure is real. Tools like Obsidian + CLI + MCP are becoming viable alternatives precisely because they offer programmatic access. Recall's content ingestion and summarization is best-in-class, but without API/MCP access, power users running agentic workflows will eventually migrate to tools that integrate into their pipelines.
Happy to beta test if you need early adopters for this.
---
T
Torsten Sandor
I was one of the first paying customers of Recall. I think I literally subscribed on day one, even emailed the team enthusiastically saying this is exactly what I was looking for.
I canceled my renewal today.
This is a MUST HAVE feature and it's been ignored for forever.
I need a different provider -- sorry.
J
Jarad Nelson
Guys, the fact that you are implementing things with 9 ratings and ignoring the second most requested feature which is still under review (which i doubt, its being ignored), something that is just crazy not to have in todays world, is a definite self-serving choice thats not in the best interests of your customer base. Respectfully, get your act together and care about what your customers actually want. Very disappointing, as this amazing app could have been beyond epic as part of a larger system, im actually depressed now. Thank you.
D
Dailen Gunter
Jarad Nelson Agreed.
With the surging popularity and rise in popularity of tools like Codex, Claude Code, and OpenClaw, Recall is on the cusp of being replaced by combining tools with something like Obsidian CLI.
I'm using Zo.computer (somewhat similar to OpenClaw) to consolidate AI tools and the lack of a MCP or API is holding things back.
n
nicole
Merged in a post:
Integrate Recall into an AI workflow, where I can ask another model a question and it checks my Recall KB first before researching the Internet?
n
nicole
So one major advantage something like NotebookLM has is the ability to digest sources in the notebook and use these to formulate the response.
So for example, I would previously use Claude to research a topic, and there would be a workflow that would then send this prompt and the Claude response to Gemini, who would then do its own research but would use my notebook as its main source, with the instruction that my notebook carried more weight because there were sources I had chosen and checked. Then Gemini would formulate a response based on not only its and Claude’s research, but also from my own notebook where I’d captured my own sources which I saw as genuine etc.
It would be great if recall replaced NotebookLM in this kind of workflow.
Sankari Nair
Merged in a post:
MCP and or API Access
G
GrumpiFied OGGVCT
this creates incredible knowledge bases to use, but it doesn't offer a clean way to access them agentically. likein a tool stack and IDE based codeing tool. i think that's the missinglink to bring this tool into the modern arena or AI solutions. this could be MCP, AND OR API access
M
Massimo Scola
fully agree - MCP would be great! I could save from Claude, Claude Code or similar straight to my knowledge
L
Luc Comeau
It's what's stopping me from uploading my entire obsidian vault
Load More
→