br8n
01Plugin Claude Code
Plugin marketplace listing 1 plugin: br8n.
Plugin Claude Code
Plugin marketplace listing 1 plugin: br8n.
Plugin Claude Code
A self-maintaining, portable personal knowledge base. Captures developer context in the background, distills it into a cross-repo graph, and asks for guidance at one seam: its schema. Skills: /br8n:pickup (where I was + cross-repo selector), /br8n:capture (save context), /br8n:search (grounded answer), /br8n:explore…
Settings file Claude Code
Git stash for your head. Saves what you were thinking when you get pulled away — branch, files, diff, and the one-line why — and replays it as a 30-second resume card. Claude Code plugin + MCP server. Local-first, MIT.
MCP server Claude CodeCodexCursor
MCP server "br8n" as configured in anthonysuherli/br8n. Launched with ${CLAUDE_PLUGIN_ROOT}/bin/br8n-mcp.
Instructions file
Claude Code instructions for anthonysuherli/br8n, covering br8n, design philosophy, architecture, what's new (br8n-specific) and what's reused from the fork.
Hook
Runs when you submit a prompt, before the agent sees it, executing preamble-inject.py. From anthonysuherli/br8n.
Hook
Runs when a session starts, executing first-run-init.py and auto-capture.py (2 commands). From anthonysuherli/br8n.
Hook
Runs when the agent finishes a response, executing session-note.py. From anthonysuherli/br8n.
Skill Claude CodeCodex
Query your cross-repo ACTIVITY knowledge graph — what you've been working on across all repos (repos touched, branches, files, work sessions, and the tasks behind them). Use when the user asks "what was I working on", "what have I touched lately", "what's connected to ", or wants a cross-repo rollup rather than a…
Skill Claude CodeCodex
Save the current workspace context to the br8n session KB as a snapshot — branch, open/cursor files, git diff stat, and a one-line hypothesis of what you're doing — AND sweep every other live Claude Code session, snapshotting each one's repo+branch in the background. Use when the user is about to switch away, says…
Skill Claude CodeCodex
Browse the auto-generated .br8n/docs/ tree distilled from your session notes — list and read the curated docs, map folders to files, and surface the most relevant ones for the user's question; pass --rebuild to force a fresh distill. Use when the user wants to read the project's living docs, find the doc covering a…
Skill Claude CodeCodex
Inspect or switch how br8n embeds text for semantic search — remote (API key) or local (keyless, on-device ONNX). Use when the user asks why search returns nothing, wants search to work without an API key, wants to stop paying for embeddings, or asks which embedding model br8n is using.
Skill Claude CodeCodex
Force the br8n gap-fill pipeline — plan → web search → crawl → extract → merge — to seed or deepen the repo+branch KB from the web, then rebuild the synopsis. Use when /br8n:pickup returns coverage='gap', when starting a KB from scratch, or when the user explicitly wants fresh online research persisted into the…
Skill Claude CodeCodex
Your cross-project personal journal — write an entry any time (insight, reflection, reference, decision) and search it back by meaning. Distinct from session notes (which are per repo+branch, written at session end); the journal is global and on-demand. Use when the user wants to jot a durable thought, "journal this"…
Skill Claude CodeCodex
Dictate what kind of session notes br8n takes for this repo+branch — show the current note policy (section template + steer), edit it inline with free-text guidance, or launch a guided wizard to co-design it. Use when the user wants to control what end-of-session notes capture ("focus on architecture", "skip dep…
Skill Claude CodeCodex
Pick up where you left off — replay the resume card for the current repo+branch, or open a cross-repo selector over every repo you've captured to and resume any of them. Use when the user returns to work, asks "where was I", "what was I doing", or wants to jump back into a different repo than the one they're in.
Skill Claude CodeCodex
Design or reshape this repo+branch's knowledge-graph schema (its ontology) — the one human-in-the-loop seam of the self-maintaining KB. Use when the user wants to set up, review, or re-design the KG schema, when a drift offer was surfaced ("your graph has drifted… reshape it?"), or proactively to check whether the…
Skill Claude CodeCodex
Ground a question in the br8n session KB, fill the gap from the web if needed, then answer — the ground → grow → answer front door. Use by default when the user asks a question against what this repo+branch session knows; reach for /br8n:pickup or /br8n:explore directly only when you want exactly one step.
Skill Claude CodeCodex
Read back your append-only activity timeline for this repo+branch — a chronological log of session notes, context captures, and journal entries. Shows recent activity (last few days) by default, with the past-week view and the full all-time scroll alongside. Pass --rebuild to force a fresh rollup now. Use when the…
Plugin Claude Code
Plugin marketplace listing 1 plugin: delapan.
Plugin Claude Code
Agentic knowledge-base engine — ground answers in maintained findings, fill gaps from the web. Skills: /delapan:resume (tap KB → preamble + coverage resume card), /delapan:search (semantic recall over findings), /delapan:explore (gap-fill from the web), /delapan:ingest (store agent findings, near-duplicates cost one…
Skill Claude CodeCodex
Show the KB's curation backlog — gap/sparse queries it was asked and couldn't answer, ranked by demand. Use when deciding what to research next, or before running explore without a topic.
Skill Claude CodeCodex
Run the gap-fill exploration pipeline (plan → search → crawl → extract → merge) when KB coverage is sparse or gap. Blocks 1–3 minutes. Requires AIGATEWAYAPIKEY and TAVILYAPIKEY in the plugin's .env.
Skill Claude CodeCodex
Research a topic using your own web search and reading, then persist the findings to the KB via delapanaddfindings. Use INSTEAD of /delapan:explore when you can search the web yourself — novel findings cost no LLM call, only near-duplicates do. Falls back to /delapan:explore for unattended or deployed runs.