Use a portable Skill or structured MCP tools, backed by the same scoped Public V1 API.
Choose the interface your agent already understands. The API behavior and permission model stay consistent.
A portable instruction bundle with a self-contained CLI. It gives an agent workflows, safety rules, and direct Public V1 API access without an MCP runtime.
Best for Codex, Claude Code, OpenCode, and other agents that can run local commands.
A published server exposing structured tools with validated inputs. Your client discovers the available operations and calls them through the MCP protocol.
Best for desktop assistants and agent hosts with native MCP support.
Both adapters use the same Public V1 API client, bearer authentication, scoped abilities, and normalized errors.
Move from retrieval to durable knowledge without teaching every agent the API by hand.
Search semantically, inspect the matching chunk, then open the source document before answering.
Save a meeting recap, research note, or session summary as a draft for human review.
Create folders and tags, merge duplicates, and place documents where they remain discoverable.
Upload and download files, inspect share state, and reserve sensitive changes for explicit requests.
Paste one short prompt into your agent. It can identify the right adapter and guide you without exposing your API key.
The prompt points to the current machine-readable guide, so the page stays concise as installation details evolve.
Help me install MarkNest for this agent.
Read https://marknest.net/llms.txt and follow the Agent Integration guide...Help me install MarkNest for this agent.
Read https://marknest.net/llms.txt and follow the Agent Integration guide.
First determine whether this client supports Skills or MCP, then ask which operating system I use.
Guide me one step at a time. Keep my API key private, ask before changing configuration files, and finish with a read-only verification.Create a scoped key, install the adapter that matches your agent, then run a real read before enabling writes.
Open Settings, create an API key, and grant only the abilities the workflow needs.
Open API key settingsDownload and verify the Skill ZIP, then extract its marknest folder into your agent skill directory. For MCP, add the npm server to your client config.
List documents or run semantic search first. Add write abilities only when the workflow calls for them.
curl -LO https://marknest.net/downloads/marknest-skill-0.2.0.zip
curl -LO https://marknest.net/downloads/marknest-skill-0.2.0.zip.sha256
shasum -a 256 -c marknest-skill-0.2.0.zip.sha256
unzip marknest-skill-0.2.0.zip -d <agent-skill-directory>export MARKNEST_API_KEY="mk_your_api_key"
node scripts/marknest.mjs search semantic \
--query "launch notes"
node scripts/marknest.mjs documents create \
--title "Launch recap" \
--content "# Launch recap" \
--status draft{
"mcpServers": {
"marknest": {
"command": "npx",
"args": ["-y", "@marknest/mcp-server@0.2.0"],
"env": {
"MARKNEST_API_KEY": "mk_your_api_key"
}
}
}
}The integration separates routine reads from actions that can publish, overwrite, or remove knowledge.
Search and inspect context before making claims or proposing edits.
New summaries and notes default to reviewable drafts whenever the workflow allows it.
Deletion, public sharing, and overwrite-like updates require explicit user intent.
Agent integrations and scoped API keys are available on Pro and Max plans.