Combining with srag

srag is a separate tool — semantic code search across all your indexed repos via MCP. It’s complementary to second-brain: where second-brain holds curated prose (standards, decisions, lessons), srag holds every implementation pattern you’ve ever written.

Together they answer two different questions Claude needs to answer for any non-trivial task:

Question	Tool
”What’s our standard / have we hit this before?”	brain_recall
”How have I implemented this pattern before?”	search_code (srag)

Install srag

git clone https://github.com/wrxck/srag.git
cd srag && ./install.sh
srag setup     # interactive: scans + indexes your projects

The installer wires srag’s MCP into Claude Code automatically.

Combined query pattern

The plugin’s _discipline skill (when both tools are present) recommends this flow for any non-trivial task:

1. brain_recall(query) — what are the relevant standards / lessons / decisions?
2. search_code(query) (srag) — are there existing implementations to reuse?
3. Synthesise:

   Per Trilium Standards / X and srag found 3 prior implementations in fleet/foo.ts, macpool/bar.ts, … — I’ll follow that pattern.

If srag returns nothing, that’s a signal the relevant project may not be indexed yet. Suggest srag index <path> to the user.

Why two tools

• second-brain is curated — every note is human-reviewed (or human-approved-from-Claude’s-suggestion). Trust is high.
• srag is automated — it indexes everything; quality varies with the underlying code. Trust is file-by-file.
• The discipline rule is the same in both cases: cite the source. “Per fleet/src/core/secrets-rotation.ts:142…” is just as accountable as “Per Trilium Standards/X…”.

Other layers

If you have the trilium-mcp (Java) server installed too, you also get the 38 raw ETAPI tools for direct note CRUD. The brain MCP’s high-level brain_* tools handle 95% of cases; trilium-mcp is the escape hatch for raw operations (bulk fixes, attribute manipulation, etc.).