bettercallclaude

BetterCallClaude transforms legal research, case strategy, and document drafting for Swiss lawyers. It provides deep integration with Swiss legal databases, multi-lingual analysis (DE/FR/IT/EN), and built-in Anwaltsgeheimnis (attorney-client privilege) protection -- 21 agents, 30 commands, 17 skills, and 10 MCP servers covering BGE/ATF/DTF precedent research, litigation strategy, adversarial analysis, legal drafting, citation verification, document intelligence, and CAS/TAS sports arbitration across all 26 Swiss cantons.

Claude Code CLI users: this repository is Cowork Desktop only. The CLI version is at fedec65/bettercallclaude-cli.

---

Overview

BetterCallClaude provides a structured methodology for handling legal work with AI assistance. The framework consists of five interconnected phases.

---

What's New in v4.11.9

v4.11.9 — Every agent runs on the model you chose. All 21 agents previously pinned to fixed tiers (opus for the orchestrator and judicial analyst, haiku for the citation specialist and summarizer, sonnet for the rest) now declare model: inherit and follow the model selected in your main conversation — cost, quality, and latency track your explicit choice end-to-end.

What's New in v4.11.8

v4.11.8 — /doctor now catches a broken agent route. In v4.11.5 an outage left plugin agents unable to reach any connector ("No such tool available") while /doctor reported everything green, because doctor only tested the main-session route. /doctor now also dispatches the plugin's citation-specialist agent for a one-shot validate_citation call and reports "Route agent: OK" — or, if the agent route is broken, says so plainly with update instructions. A new CI check additionally enforces that every MCP tool in agent/command/skill whitelists is registered under both naming conventions hosts use, so this regression class cannot silently return.

What's New in v4.11.5

v4.11.5 — Ollama server actually starts on Cowork Desktop. The v4.11.4 fallback never ran: Cowork's host bridge drops any server whose .mcp.json entry references ${user_config.*} (user_config is not supported on the desktop host bridge; dropping server), so the whole ollama server was rejected at load. The template and the dead ollama_host plugin setting are removed — the bundled server now starts and talks to your local Ollama daemon at http://localhost:11434 directly. Update the plugin, restart Cowork, then run /bettercallclaude:setup to see the ollama server connected.

v4.11.4 — Ollama connector fixed on Cowork Desktop. Since v4.3.0 the bundled ollama server received an empty ollama_host setting on Cowork (the app exposes no userConfig settings UI), which broke every local translation/summarisation call with Failed to parse URL from /api/version. The server now falls back to the default http://localhost:11434 automatically — reinstall or update the plugin and no action is needed.

v4.11.3 — Workflow IDs that survive Cowork restarts. Cowork Desktop wipes its sandbox filesystem on restart, which used to reset your generated workflow ID. The workflow commands now also read your ID from a BetterCallClaude workflow user ID: … line in Settings → General → Instructions for Claude — one line to paste once, stored by the app, permanent. The generation message tells new users how.

v4.11.2 — Server-enforced unique workflow namespaces. The new claim_user_id tool on the workflows-ch server reserves your User ID in the database (UNIQUE constraint), so no two users can ever share a workflow namespace. Generated bcc-… IDs are claimed before they're saved, with automatic retry on collision.

v4.11.1 — Private workflow namespaces for everyone. Custom workflows are now stored under a unique per-user ID: if you don't set the User ID plugin setting, a random bcc-… ID is generated once and saved to ~/.betterask/config.yaml. No shared default namespace anymore.

v4.11.0 — Custom workflows. Build your own multi-agent pipelines once and reuse them: /bettercallclaude:create-workflow interviews you, validates the agent chain server-side (compatible hand-offs only), and saves it; /bettercallclaude:workflow then lists your saved workflows next to the built-in templates and runs them with the same engine. Backed by the new workflows-ch MCP server (30 commands, 10 MCP servers total).

Also in recent releases — v4.10.0 legal wayfinder: /legal-chart decomposes a matter into a decision map — destination, decisions so far, fog, out-of-scope — with one ticket per open decision. /legal-way then works the map ticket by ticket (research, grilling, prototype, task) until the route to the deliverable is clear and hands off to execution. A fog check in /briefing routes oversized matters to the chart instead of forcing a static execution plan.

Also in recent releases — v4.9.6 privacy fix: the Anwaltsgeheimnis PreToolUse hook silently failed on plugin paths containing spaces (e.g. a user name with a space) — privilege detection was off with no error shown. The hook command now quotes the plugin root correctly, with regression guards (standalone tests + a CI check). If you handle privileged client content, update.

Also in recent releases — v4.9.5 /legal-timeline: build a sourced case chronology from your case documents: every event carries its document and locus, contested facts and date conflicts are made visible (never silently resolved), evidentiary gaps of 30+ days are flagged, and procedural deadlines are computed via MCP with cantonal holiday calendars. Three outputs (Markdown table, interactive HTML, Word export) under bcc-output/timeline/.

Content counts: 21 agents, 30 commands, 17 skills, 10 MCP servers in .mcp.json (8 remote HTTP on mcp.bettercallclaude.ch + swiss-caselaw SSE on mcp.opencaselaw.ch + ollama local STDIO).

Full changelog →

Cowork Desktop dedicated release -- This repository is exclusively for Claude Cowork Desktop. The Claude Code CLI version is at fedec65/bettercallclaude-cli.

• HTTP-only transport: 9 of 10 MCP servers connect via mcp.bettercallclaude.ch / mcp.opencaselaw.ch -- no local Node.js build required for those
• Local STDIO server (ollama): bundled and only touches http://localhost:11434 for privacy-routed translation/summarisation
• Simplified setup: /setup checks connectivity only -- no transport switching needed in Cowork

---

Installation

Full installation guide with screenshots: BetterCallClaude Tutorial →

1. In Cowork, click Customize > Browse plugins > Personal > + > Add marketplace from GitHub
2. Enter fedec65/bettercallclaude and click Sync
3. Click Install on the BetterCallClaude card

MCP servers connect automatically via HTTP. No Node.js, no local setup, no API keys required.

---

Commands

Command	Description
/bettercallclaude:legal	Intelligent gateway -- analyzes intent, routes to the appropriate specialist agent, and manages multi-step legal workflows. Use --refine to transform vague queries first.
/bettercallclaude:refine	Transform vague legal queries into structured prompts through Socratic dialogue. Recommends optimal workflows and introduces Swiss legal terminology.
/bettercallclaude:research	Search Swiss legal precedents and compile research memoranda. Supports BGE/ATF/DTF databases, doctrine references, and cross-jurisdictional analysis.
/bettercallclaude:strategy	Develop litigation strategy with risk assessment, cost-benefit analysis, and procedural pathway evaluation.
/bettercallclaude:draft	Draft Swiss legal documents including contracts, court briefs, legal opinions, and memoranda with proper citation formatting.
/bettercallclaude:cite	Verify and format Swiss legal citations across all four national languages (BGE/ATF/DTF formats).
/bettercallclaude:validate	Validate Swiss legal citations in bulk -- check format, existence, and cross-language consistency.
/bettercallclaude:precedent	Search and analyze BGE/ATF/DTF precedents with precedent chain tracking and evolution analysis.
/bettercallclaude:federal	Analyze a legal question under federal Swiss law (ZGB, OR, StGB, BV, and related federal statutes).
/bettercallclaude:cantonal	Analyze a legal question under cantonal law for a specific canton.
/bettercallclaude:adversarial	Run three-agent adversarial analysis -- advocate builds the case, adversary challenges it, judicial analyst synthesizes.
/bettercallclaude:briefing	Structured pre-execution briefing -- assembles a specialist panel, collects case context, and builds an execution plan before agents start working.
/bettercallclaude:legal-chart	Chart a big or foggy matter as a wayfinder decision map -- one ticket per open decision, planning only.
/bettercallclaude:legal-way	Work one decision ticket from a wayfinder map; hands off to execution only when every decision is made.
/bettercallclaude:workflow	Define and execute multi-agent legal workflows (due diligence, litigation prep, contract lifecycle, real estate closing), including saved custom workflows.
/bettercallclaude:create-workflow	Create a reusable custom workflow by combining agents -- interview-based: pick agents, order them, define the output; saved for future use with /workflow.
/bettercallclaude:legal-5step	Execute the 5-step Swiss legal framework end to end: intake, research, strategy, adversarial stress test, verified drafting.
/bettercallclaude:legal-goal	Define a checkable legal success condition for /legal-loop.
/bettercallclaude:legal-loop	Iterate a worker-evaluator cycle against a Goal Record until the success condition is met or a stop limit is reached.
/bettercallclaude:translate	Translate Swiss legal documents between DE, FR, IT, and EN while preserving legal terminology precision.
/bettercallclaude:doc-analyze	Analyze Swiss legal documents -- identify legal issues, extract key clauses, verify citations, assess compliance.
/bettercallclaude:legal-timeline	Build a sourced legal chronology from case documents -- provenance per event, contested-fact status, date conflicts, gaps, deadlines.
/bettercallclaude:nda-triage	Triage NDAs against Swiss law -- GREEN / YELLOW / RED using playbook thresholds; single file or batch mode.
/bettercallclaude:summarize	Consolidate multi-agent pipeline output -- deduplicate disclaimers, terminology, and citations with length control (--short/--medium/--long).
/bettercallclaude:start	First-use onboarding -- checks MCP connectivity, guides playbook creation, shows tailored usage examples.