From c7c36e34a801f5f183f8a6e6bc377357cd1f9eb6 Mon Sep 17 00:00:00 2001 From: Bill Thornton Date: Thu, 2 Jul 2026 12:17:05 -0700 Subject: [PATCH] docs(readme): reconcile tool-count contradictions (TAP-4577) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit README stated inconsistent MCP tool counts: lines 57 and 104 claimed 84 tools (42 TappsMCP + 42 DocsMCP) while the front-door line 5, architecture diagram, and package table already said 85 (43 + 42). The dependency diagram (line 1075) and structure tree (line 1207) also undercounted TappsMCP as 42. Authoritative count is code-derived: ALL_TOOL_NAMES in packages/tapps-mcp/src/tapps_mcp/server.py = 43, ALL_DOCS_TOOL_NAMES in packages/docs-mcp/src/docs_mcp/server.py = 42, total 85 — matching docs/architecture/tool-budget.md. Corrected all four stale mentions to 43 TappsMCP / 42 DocsMCP / 85 total. Co-Authored-By: Claude Opus 4.8 (1M context) --- README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 407dc1d1..403a6933 100644 --- a/README.md +++ b/README.md @@ -54,7 +54,7 @@ tapps-brain (standalone) <── tapps-core (shared infra) <── tapps-mcp ### Key highlights -- **84 deterministic MCP tools** (42 TappsMCP + 42 DocsMCP) — no LLM calls in the tool chain; same input always produces same output +- **85 deterministic MCP tools** (43 TappsMCP + 42 DocsMCP) — no LLM calls in the tool chain; same input always produces same output - **Multi-language code scoring** - Python, TypeScript/JavaScript, Go, Rust across 7 categories (complexity, security, maintainability, test coverage, performance, structure, devex) - **Documentation lookup** via Context7 and LlmsTxt providers with local caching - **Persistent shared memory** via [tapps-brain](https://github.com/wtthornton/tapps-brain) — project decisions survive across sessions. TappsMCP accesses the Dockerized brain service over HTTP via BrainBridge; use `uv run tapps-mcp memory` (42 actions) or enable `nlt-memory` for MCP-exposed recall/save/handoff tools. See the [tapps-brain repo](https://github.com/wtthornton/tapps-brain) for retrieval, decay, consolidation, and federation internals. @@ -101,7 +101,7 @@ Any MCP-capable client (Claude Code, Cursor, VS Code Copilot, Claude Desktop, cu ## Features -The platform exposes **84 MCP tools** (42 TappsMCP + 42 DocsMCP) plus workflow prompts. All tools are **deterministic** (no LLM calls in the tool chain). +The platform exposes **85 MCP tools** (43 TappsMCP + 42 DocsMCP) plus workflow prompts. All tools are **deterministic** (no LLM calls in the tool chain). ### Code quality & scoring @@ -1072,7 +1072,7 @@ tapps-core (shared infrastructure) ^ ^ | | tapps-mcp docs-mcp -(42 tools) (42 tools) +(43 tools) (42 tools) ``` **[tapps-brain](https://github.com/wtthornton/tapps-brain)** is the standalone memory service extracted from tapps-core. It runs as a Dockerized Postgres-backed service that TappsMCP clients reach over HTTP at `localhost:8080`. Storage engine, retrieval (BM25 + boosts), time-based decay, contradiction detection, consolidation, federation, and GC internals all live in the [tapps-brain repo](https://github.com/wtthornton/tapps-brain) and its README/CHANGELOG — refer there for the authoritative description so this page doesn't drift. tapps-brain has its own release cycle and test suite. @@ -1204,7 +1204,7 @@ packages/ │ ├── metrics/ # Collector, dashboard, alerts, trends, OTel export │ └── adaptive/ # Adaptive scoring, expert voting, weight distribution │ -├── tapps-mcp/ # Code quality MCP server (42 tools) +├── tapps-mcp/ # Code quality MCP server (43 tools) │ └── src/tapps_mcp/ │ ├── server.py, cli.py # Entry points and MCP server │ ├── server_*.py # Tool modules (scoring, pipeline, metrics, memory, analysis)