@standardbeagle/slop-mcp 0.12.1 → 0.13.0

package/README.md

@@ -61,15 +61,19 @@ claude mcp add slop-mcp -- npx -y @standardbeagle/slop-mcp serve
 Define MCP servers in `.slop-mcp.kdl`:
 
 ```kdl
-mcp "filesystem" {
-    command "npx" "-y" "@anthropic/mcp-filesystem"
-    args "/path/to/dir"
-}
-
 mcp "github" {
     transport "sse"
     url "https://mcp.github.com/sse"
 }
+
+mcp "jira" {
+    transport "streamable"
+    url "https://mcp.atlassian.com/v2/mcp"
+}
+
+mcp "lci" {
+    command "lci" "mcp"
+}
 ```
 
 ## The 8 Meta-Tools
@@ -85,14 +89,60 @@ mcp "github" {
 | `slop_reference` | Browse SLOP built-in functions |
 | `slop_help` | Get detailed help for a SLOP function |
 
+## Examples
+
+### Cross-MCP orchestration with SLOP
+
+Chain tools across multiple MCPs in a single `run_slop` call — intermediate results stay out of the agent's context:
+
+```python
+# Create Jira tasks from unread emails matching a filter
+emails = gmail.search_messages(query: "label:action-needed is:unread")
+for email in emails {
+    jira.create_issue(
+        project: "OPS",
+        summary: email["subject"],
+        description: format("From: {}\n\n{}", email["from"], email["snippet"]),
+        issue_type: "Task"
+    )
+    gmail.modify_message(id: email["id"], remove_labels: ["UNREAD"])
+}
+emit(created: len(emails))
+# Agent sees only: {"created": 4}
+```
+
+```python
+# Index codebase structure into persistent memory for future sessions
+results = lci.search(query: "public API endpoints")
+endpoints = results
+    | map(|r| {"path": r["file"], "name": r["symbol"], "kind": r["kind"]})
+    | filter(|r| r["kind"] == "function")
+mem_save("project", "api_endpoints", endpoints,
+    description: "Public API endpoint inventory")
+emit(indexed: len(endpoints))
+```
+
+```python
+# Generate a visual report from code analysis
+stats = lci.search(query: "struct")
+by_package = stats
+    | map(|r| r["file"] | split("/") | first())
+    | group_by(|pkg| pkg)
+chart_data = by_package
+    | items()
+    | map(|pair| {"label": pair[0], "value": len(pair[1])})
+    | sorted(|a, b| b["value"] - a["value"])
+banana.create_chart(type: "bar", title: "Structs by Package", data: chart_data)
+```
+
 ## Features
 
 - **Progressive discovery** — agents find tools via `search_tools`, not by loading everything upfront
-- **SLOP scripting** — chain tool calls and process results in a single `run_slop` call
+- **SLOP scripting** — chain tool calls across MCPs and process results in a single `run_slop` call
 - **Lazy connections** — MCP servers connect asynchronously with tool metadata caching
 - **Persistent memory** — disk-backed `mem_save`/`mem_load`/`mem_search` across sessions
 - **Three-tier config** — project-local > project > user config merging with KDL
-- **OAuth support** — browser-based auth for MCPs like Figma, GitHub, Linear
+- **OAuth support** — browser-based auth for MCPs like Figma, GitHub, Linear, Jira
 - **All transports** — stdio, SSE, and streamable HTTP
 
 ## Documentation

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@standardbeagle/slop-mcp",
-  "version": "0.12.1",
+  "version": "0.13.0",
   "description": "MCP orchestrator — connect unlimited Model Context Protocol servers through 8 meta-tools. Progressive tool discovery keeps your AI agent's context window small.",
   "main": "index.js",
   "bin": {