Obsidian + Wiki + AGNT Knowledge Playbook
Wire your Obsidian vault into AGNT's knowledge layer — audit links, persist agent memory, and query the graph.
Why this playbook
AGNT agents produce knowledge every heartbeat — decisions, learnings, signals, venue data. Without a structured knowledge layer, this information scatters across conversations and logs. Obsidian gives you a searchable, linked, human-readable view of everything your fleet knows.
This playbook connects Obsidian to AGNT's memory system via MCP. Step 1 audits your existing wiki for broken links and stale content. Step 2 configures persistent memory writing so agents automatically log learnings. Step 3 teaches you to query the knowledge graph from any Claude Code session.
The result: a self-maintaining wiki that grows with your agent fleet, queryable from any IDE or terminal session.
Prerequisites
- Obsidian installed with the Local REST API plugin enabled
- An existing vault (even an empty one works)
- Claude Code with MCP server support
- AGNT MCP server configured (agnt-wiki-fs or obsidian adapter)
Input requirements
| Input | Type | Required | Description |
|---|---|---|---|
| Obsidian vault path | directory path | Yes | Absolute path to your Obsidian vault root. The MCP server needs filesystem access to this directory. |
| MCP server type | string | Yes | 'obsidian' (REST API adapter) or 'agnt-wiki-fs' (direct filesystem). REST API is recommended for cross-device sync compatibility. |
| PARA folder structure | boolean | No | Whether to use Tiago Forte's PARA method (Projects, Areas, Resources, Archive) for organizing knowledge. Default: yes. |
Step-by-step workflow
Audit existing wiki for broken links and stale content
Open promptStart here even if your vault is empty — the auditor sets a baseline and creates the folder structure your agents will write into. If you have existing content, it identifies orphan notes, broken wikilinks, and pages that haven't been updated in 30+ days.
The audit produces a structured report with fix-or-delete recommendations for every issue found. Run this monthly to keep your knowledge base healthy.
Configure persistent agent memory writing
Open promptThis prompt configures the MCP memory layer so your agents write structured notes to the vault after every significant action. Each note uses YAML frontmatter for metadata and markdown for content, following the PARA folder convention.
You'll configure which events trigger a write (heartbeat completions, errors, decisions, learnings) and what format the notes use. The prompt generates the MCP tool definitions and the hook configuration.
Query the knowledge graph from Claude Code
Open promptThe final step teaches you to query your accumulated knowledge from any Claude Code session. You'll set up search patterns, tag-based retrieval, and cross-reference queries that pull context from your wiki into active coding sessions.
This is where the investment pays off: every future conversation can draw on the full history of what your agents have learned.
Expected outputs
Wiki audit report
Structured report of broken links, orphan notes, stale content, and folder structure recommendations.
Produced by step 1MCP memory configuration
MCP tool definitions, hook config, and PARA folder templates for automated agent memory writing.
Produced by step 2Query templates
Reusable search patterns and MCP tool calls for querying your knowledge graph from any Claude Code session.
Produced by step 3Tool requirements
- Claude Code with MCP server access
- Obsidian vault with REST API plugin enabled
- AGNT MCP server (agnt-wiki-fs or obsidian)
Troubleshooting
MCP server can't connect to Obsidian REST API
Agent writes create duplicate notes
Graph queries return no results
Share as social post
Obsidian + Wiki + AGNT Knowledge Playbook 3 steps, 45-60 minutes. Wire your Obsidian vault into AGNT's knowledge layer — audit links, persist agent memory, and query the graph. https://agntdot.com/playbooks/obsidian-knowledge-playbook
236 / 280 chars