Solve stale context 800× faster.
Live markdown truth for team knowledge graphs, Kanban boards, documents and CRMs.
Help me install Native: https://github.com/withnative/native-pluginHelp me install Native MCP and use its quickstart: https://plugin.withnative.ai/mcpMore signal, less noise
Health warnings on stale context. Agents and teammates kept in-sync.
Fast context, fewer mistakes
Efficient queries, so your markdown context scales. Focused retrieval, less token bloat.
SQL queries are 36,000× faster
“What active work is waiting on something unfinished?”
- query_sql tool call
SELECT w.name, b.name FROM links JOIN records …View query
SELECT w.name, b.name FROM links l JOIN records w ON w.id = l.source_id JOIN records b ON b.id = l.target_id WHERE l.relationship IN ('depends_on', 'blocked_by') AND w.type = 'WorkItem' AND w.lifecycle IN ('open', 'in_progress', 'blocked') AND b.lifecycle IN ('open', 'in_progress', 'blocked'); - using a CLI processor, awk
awk -f join.awk *.md - markdown search
for f in $(find . -name '*.md'); do …View query
cd "Aldermere Ferry Co" # join: what active work is waiting on something unfinished? # every edge is a bare [[Name]] — no type, no direction — so a # mention reads as a dependency and depends_on cannot be told # from part_of or blocks for f in $(find . -name '*.md'); do grep -qE '^type: WorkItem' "$f" || continue grep -qE '^lifecycle: (open|in_progress|blocked)' "$f" || continue while IFS= read -r link; do t=$(grep -rl "^title: ${link}$" . | head -1) [ -n "$t" ] && grep -qE '^lifecycle: (open|in_progress|blocked)' "$t" \ && echo "$(basename "$f") waits on $link" done < <(grep -oE '\[\[[^]]+\]\]' "$f" | tr -d '[]') done
Realtime truth, simple coordination
Simple and fast coordination. New context is live, instantly, for your AI and teammates. Get your time and headspace back.
Easy to start. Yours to own.
Context that’s uniquely yours. Choose where you store it, how it’s modelled, who gets to see it and how you render it.
| Capability | Native | Obsidian | Notion | Google Drive | Vendor memory | Enterprise search |
|---|---|---|---|---|---|---|
| Open-source engine | yes | no | no | no | no | no |
| Transparent markdown content | yes | yes | no | no | no | no |
| Real-time cloud context | yes | partial | yes | yes | yes | yes |
| Run locally or on your server | yes | yes | no | no | no | no |
| Any agent or model | yes | yes | yes | yes | no | yes |
| Granular sharing controls | yes | partial | yes | yes | partial | yes |
| Custom data model | yes | partial | yes | no | no | no |
| Agents can rewrite your UI | yes | partial | no | no | no | no |
FAQs
What is Native?
Native gets you and your team better results from AI through live, shared context.
It means that you get:
– Simple and fast coordination because new context doesn't need hand-offs
– Better results from AI, with fresher context and less wrong-footing
– More time and headspace to think big about goals and priorities
It levels up your markdown content a shared, realtime workspace. It's powered by structured SQL queries that run up to 36,000 faster than file-search alternatives.
It's fast and easy to start. Our MCP server is a one-line installation, and there are plugins for both Claude and ChatGPT / Codex.
Native's engine is 100% open-source. You can run in our cloud with our Free Forever plan, or self-host.
How is Native different to a markdown vault or repo?
Native gives your markdown content more velocity, reach and power. It's great for individuals and teams leaning heavily into AI and agents.
– Indexed answers across linked records. Native uses SQL to query structured data and relationships, so agents can find the relevant work quickly.
– Live context without push/pull. Native is great for cloud-first context. There's no manual refresh steps required. Coordination is fast and easy.
– Simple, granular sharing. Native lets you easily share context with teammates, or hide what's sensitive. You're in control.
– Advisory agent feedback. Native tells your AI agents if they're fetching stale content, and re-points them in the right direction.
– Interactive, adaptable UI. Native works as a realtime backend for whatever UI you want. Use ours or write your own.
We recommend Native's cloud, but you can bring your own cloud or even run Native locally. You're in full control of your context.
How is Native different to software like Notion?
Native gives more freedom and power to people using AI agents, and was built ground-up for this paradigm.
– Better grounding for agents with advisory feedback. Native tells your AI agents if they're fetching stale content, and re-points them in the right direction.
– Ownership of your key AI asset: your context. You decide where your data is stored. Our cloud is easiest, but you can self-host or even run it locally.
– Make it yours with a modifiable UI and engine. Native's core is open-source, and works as a realtime backend for custom UI. You, or your agents, are free to adapt it.
What's the Free Forever plan? What's the catch?
Native is open-source. We are debuting our early alpha release at the Long Horizon Agents hackathon!
Our Free Forever plan is a perk for early users, with a minimum of 2GB storage planned. (We need more data to figure out other dimensions.)
You can always export and self-host if you find cheaper, more effective storage.
There's no formal catch or condition.
We appreciate encouragement and support through GitHub stars!
How are the 800× and 36,000× figures calculated?
We used a demo workspace, which models tasks and dependencies amongst other pieces of context.
Then, we wrote queries to answer the question: "Which active tasks depend on unfinished work?"
(We didn't bother using LLM inference, since this type of question ought to be a deterministic query that is cheap and fast.)
– In normal markdown, this can only be answered deterministically with a file search script. This took about 699ms.
– We tested awk for a best-case, which is a more efficient CLI text processor. This took about 15ms.
– Native uses SQLite. An indexed SQLite query took 0.019ms, which is 800x as fast as awk and 36,000x as fast as the file search script.
More complex queries, involving multiple joins of structured data, would likely show even larger speed savings.
Why are you building this and making it open-source?
In the age of AI, we believe that a key freedom is the right to own your context.
To be a true freedom, it must be structurally guaranteed, rather than dependent on provider discretion.
Open-sourcing Native is our commitment to that key freedom.
Read more about our mission here.
Who's behind Native?
Native's open-source software is written by one core developer with some part-time contributors.
If you're interested in our work, come say hello to Richard at the hackathon!
(If you're reading this subsequently: please contact hello@withnative.ai)