chughtapan · chughtapan · Oct 8, 2025 · Oct 8, 2025
diff --git a/docs/benchmarks.md → docs/evals.md b/docs/benchmarks.md → docs/evals.md
@@ -1,6 +1,6 @@
-# Running Benchmarks
+# Running Evals
 
-<em class="wags-brand">wags</em> includes evaluation support for the [Berkeley Function Call Leaderboard (BFCL)](https://gorilla.cs.berkeley.edu/leaderboard.html), enabling systematic testing of LLM function calling capabilities across multi-turn conversations.
+Here's how to run BFCL multi-turn evaluations with <span class="wags-brand">wags</span>.
 
 ## Setup
 
@@ -33,7 +33,7 @@ git submodule update --remote
 # Run specific test
 .venv/bin/pytest 'tests/benchmarks/bfcl/test_bfcl.py::test_bfcl[multi_turn_base_121]'
 
-# Run test category
+# Run test category (multi_turn_base, multi_turn_miss_func, multi_turn_miss_param, multi_turn_long_context)
 .venv/bin/pytest tests/benchmarks/bfcl/test_bfcl.py -k "multi_turn_miss_func"
 ```
 
@@ -69,27 +69,7 @@ Validate existing logs without running new tests:
 .venv/bin/pytest tests/benchmarks/bfcl/test_bfcl.py --validate-only --log-dir outputs/experiment1/raw
 ```
 
-## Test Categories
-
-- **multi_turn_base**: Standard multi-turn function calling (800 tests)
-- **multi_turn_miss_func**: Tests handling of missing function scenarios
-- **multi_turn_miss_param**: Tests handling of missing parameters
-- **multi_turn_long_context**: Context window stress tests with overwhelming information
-- **Memory tests**: Tests with key-value, vector, or recursive summarization backends
-
-
-## Developer Guide
-
-1. **Discovery**: pytest collects tests from `loader.find_all_test_ids()`
-2. **Setup**: Creates MCP servers wrapping BFCL API classes using `uv run python`
-3. **Execution**: Runs multi-turn conversations with FastAgent
-4. **Serialization**: Saves complete message history to `complete.json`
-5. **Extraction**: Extracts tool calls from JSON (preserves what FastAgent drops)
-6. **Validation**: Uses BFCL validators to check correctness
-7. **Result**: Pass/fail based on `validation["valid"]`
-
 ## Further Reading
 
-- **Test organization and patterns**: See [tests/README.md](../tests/README.md)
 - **BFCL leaderboard**: Visit [gorilla.cs.berkeley.edu](https://gorilla.cs.berkeley.edu/leaderboard.html)
 - **Official BFCL repository**: [github.com/ShishirPatil/gorilla](https://github.com/ShishirPatil/gorilla)
diff --git a/docs/index.md b/docs/index.md
@@ -10,12 +10,20 @@ The <span class="wags-brand">wags</span> toolkit is based on state-of-the-art re
 
 <div class="grid cards" markdown>
 
-- **[Getting Started](quickstart.md)**
+- **[Quick Start](quickstart.md)**
 
-    Install <span class="wags-brand">wags</span> and create your first <span class="wags-brand">wags</span> enhanced server in minutes. See [`servers/github/`](https://github.com/chughtapan/wags/tree/main/servers/github) for a complete example.
+    Connect to existing <span class="wags-brand">wags</span> servers and start using enhanced MCP features immediately.
+
+- **[Onboarding New Servers](onboarding.md)**
+
+    Create your first <span class="wags-brand">wags</span> enhanced server with middleware. See [`servers/github/`](https://github.com/chughtapan/wags/tree/main/servers/github) for a complete example.
 
 - **[Middleware](middleware/overview.md)**
 
-    Learn about available middleware and how to use them
+    Learn about available middleware and how to use them.
+
+- **[Running Evals](evals.md)**
+
+    Execute BFCL benchmarks to evaluate agent performance with <span class="wags-brand">wags</span> servers.
 
 </div>
diff --git a/docs/middleware/todo.md b/docs/middleware/todo.md
@@ -11,40 +11,29 @@
 proxy = create_proxy(server, enable_todos=True)
 ```
 
-## Example Workflow
+## How it works?
 
-```python
-# Agent receives task: "Build project and fix errors"
-
-# 1. Agent calls TodoWrite
-TodoWrite(todos=[
-    {"content": "Build project", "status": "pending"},
-    {"content": "Fix errors", "status": "pending"}
-])
-
-# 2. Agent starts first task
-TodoWrite(todos=[
-    {"content": "Build project", "status": "in_progress"},
-    {"content": "Fix errors", "status": "pending"}
-])
-
-# 3. Agent runs build tool, finds 3 errors
-
-# 4. Agent updates todos
-TodoWrite(todos=[
-    {"content": "Build project", "status": "completed"},
-    {"content": "Fix error in utils.py", "status": "in_progress"},
-    {"content": "Fix error in api.py", "status": "pending"},
-    {"content": "Fix error in models.py", "status": "pending"}
-])
-
-# 5. Agent fixes each error, updating status after each one
-# ... continues until all completed
-```
+When todo integration is enabled, the target MCP server is provided TodoWrite tools which help track the detailed tasks to be done and the current progress. Additionally, detailed instructions are provided to break down task into actionable steps, update status before and after each task, maintain exactly one task as `in_progress`, mark completed immediately after finishing, etc.
 
-## How it works?
+For example, when an agent receives the task "Build project and fix errors":
+
+1. Agent calls `TodoWrite` to create initial todos:
+   - "Build project" (pending)
+   - "Fix errors" (pending)
+
+2. Agent starts first task by updating status to `in_progress`
+
+3. Agent runs build tool, finds 3 errors
+
+4. Agent updates todos to reflect discovered errors:
+   - "Build project" (completed)
+   - "Fix error in utils.py" (in_progress)
+   - "Fix error in api.py" (pending)
+   - "Fix error in models.py" (pending)
+
+5. Agent fixes each error, updating status after each one until all completed
 
-When todo integration is enabled, the target MCP server is provided TodoWrite tools which help track the detailed tasks to be done and the current progress. Additionally, detailed instructions are provided to break down task into actionable steps, update status before and after each task, maintain exactly one task as `in_progress`, mark completed immediately after finishing, etc. See `src/wags/middleware/todo.py` for the full instruction text.
+See `src/wags/middleware/todo.py` for the full instruction text.
 
 **Note:** Instructions from proxy server must be included in the agent prompt. For `fast-agent` the `{{serverInstructions}}` macro enables this feature.
 

diff --git a/docs/onboarding.md b/docs/onboarding.md
@@ -0,0 +1,98 @@
+# Onboarding New Servers
+
+Learn how to create a <span class="wags-brand">wags</span> proxy server that wraps existing MCP servers with middleware.
+
+## Prerequisites
+
+- <span class="wags-brand">wags</span> installed (see [Quick Start](quickstart.md) for installation)
+- Basic understanding of [MCP (Model Context Protocol)](https://modelcontextprotocol.io/docs/getting-started/intro)
+- An existing MCP server to work with
+
+## Creating a <span class="wags-brand">wags</span> Proxy Server
+
+<span class="wags-brand">wags</span> provides the `quickstart` command to generate proxy servers that wrap existing MCP servers with middleware.
+
+!!! tip "Complete Example Available"
+    The complete implementation for the [GitHub MCP Server](https://github.com/github/github-mcp-server) is in `servers/github/`.
+
+### Step 1: Prepare Your MCP Server Configuration
+
+Create a configuration file that describes your MCP server. Save it as `config.json`:
+
+```json title="config.json"
+--8<-- "snippets/quickstart/config.json"
+```
+
+### Step 2: Generate the Proxy Server
+
+Use the `quickstart` command to generate middleware handlers and main file:
+
+```bash
+# Generate both handlers and main files
+wags quickstart config.json
+
+# Or with custom file names
+wags quickstart config.json \
+  --handlers-file github_handlers.py \
+  --main-file github_proxy.py
+```
+
+### Step 3: Add Middleware Decorators
+
+Edit the generated handlers file to add middleware decorators:
+
+```python linenums="1" title="handlers.py"
+--8<-- "snippets/quickstart/handlers.py"
+```
+
+### Step 4: Attach Middleware to your MCP Server
+
+The automatically generated main.py includes (commented) code to attach <span class="wags-brand">wags</span> middleware to your MCP server. You should edit the file to uncomment the middleware you need:
+
+```python linenums="1" title="main.py"
+--8<-- "snippets/quickstart/main.py"
+```
+
+### Step 5: Run Your Proxy Server
+
+```bash
+# Run directly
+python main.py
+
+# Or use wags start-server
+wags start-server servers/your-server
+```
+
+Your proxy server is now running!
+
+### Step 6 (Optional): Add to Shared Configuration
+
+To use your server with `wags run`, add it to `servers/fastagent.config.yaml`:
+
+```yaml
+mcp:
+  servers:
+    your-server:
+      transport: stdio
+      command: wags
+      args:
+        - start-server
+        - servers/your-server
+      env:
+        API_KEY: ${YOUR_API_KEY}
+      roots:
+        - uri: https://example.com/allowed
+          name: "Allowed Resources"
+```
+
+Now you can connect to your server with:
+
+```bash
+wags run --servers your-server
+```
+
+## Learn More
+
+- **[Middleware Overview](middleware/overview.md)** - Understand how middleware works
+- **[Roots](middleware/roots.md)** - Access control with URI templates
+- **[Elicitation](middleware/elicitation.md)** - Parameter review and collection
diff --git a/docs/quickstart.md b/docs/quickstart.md
@@ -1,17 +1,15 @@
 # Quick Start Guide
 
-Get up and running with <span class="wags-brand">wags</span> in just a few minutes. This guide will walk you through installation and creating a proxy server with middleware for existing MCP servers.
+Get started with <span class="wags-brand">wags</span> by connecting to existing wags servers.
 
 ## Prerequisites
 
 - Python 3.13.5 or higher
-- [`uv` package manager](https://docs.astral.sh/uv/getting-started/installation/) (recommended) or `pip`
-- Basic understanding of [MCP (Model Context Protocol)](https://modelcontextprotocol.io/docs/getting-started/intro)
-- An existing MCP server to work with
+- [`uv` package manager](https://docs.astral.sh/uv/getting-started/installation/)
 
 ## Installation
 
-> ⚠️ **Warning**: <span class="wags-brand">wags</span> is based on ongoing research and is under active development. Features and APIs may change.
+> ⚠️ **Warning**: <span class="wags-brand">wags</span> is based on ongoing research and is under active development. Features and APIs may change. Some experimental MCP features are only supported in our fork of [fast-agent](https://github.com/chughtapan/fast-agent) included with <span class="wags-brand">wags</span>.
 
 ```bash
 # Clone the repository
@@ -38,61 +36,24 @@ WAGS version 0.1.0
 FastMCP version x.x.x
 ```
 
-## Creating a <span class="wags-brand">wags</span> Proxy Server
+## Getting Started
 
-<span class="wags-brand">wags</span> provides the `quickstart` command to generate proxy servers that wrap existing MCP servers with middleware.
-
-!!! tip "Complete Example Available"
-    The complete implementation for the [GitHub MCP Server](https://github.com/github/github-mcp-server) is in `servers/github/`.
-
-### Step 1: Prepare Your MCP Server Configuration
-
-Create a configuration file that describes your MCP server. Save it as `config.json`:
-
-```json title="config.json"
---8<-- "snippets/quickstart/config.json"
-```
-
-### Step 2: Generate the Proxy Server
-
-Use the `quickstart` command to generate middleware handlers and main file:
+The easiest way to connect to wags servers is using the `wags run` command:
 
 ```bash
-# Generate both handlers and main files
-wags quickstart config.json
-
-# Or with custom file names
-wags quickstart config.json \
-  --handlers-file github_handlers.py \
-  --main-file github_proxy.py
-```
+# Connect to all configured servers
+wags run
 
-### Step 3: Add Middleware Decorators
+# Connect to specific servers only
+wags run --servers github
 
-Edit the generated handlers file to add middleware decorators:
-
-```python linenums="1" title="handlers.py"
---8<-- "snippets/quickstart/handlers.py"
-```
-
-### Step 4: Attach Middleware to your MCP Server
-
-The automatically generated main.py includes (commented) code to attach <span class="wags-brand">wags</span> middleware to your MCP server. You should edit the file to uncomment the middleware you need:
-
-```python linenums="1" title="main.py"
---8<-- "snippets/quickstart/main.py"
-```
-
-### Step 5: Run Your Proxy Server
-
-```bash
-python main.py 
+# Use a different model
+wags run --model claude-3-5-sonnet-20241022
 ```
 
-Your proxy server is now running! 
+`wags run` invokes fast-agent with a configuration file ([`servers/fastagent.config.yaml`](https://github.com/chughtapan/wags/blob/main/servers/fastagent.config.yaml)) and basic instructions ([`src/wags/utils/agent_instructions.txt`](https://github.com/chughtapan/wags/blob/main/src/wags/utils/agent_instructions.txt)), and connects to all servers by default. You can configure which servers to connect to using the `--servers` flag or create your own configuration and instruction files - see the [fast-agent documentation](https://github.com/chughtapan/fast-agent) for more details.
 
-## Learn More
+## Next Steps
 
-- **[Middleware Overview](middleware/overview.md)** - Understand how middleware works
-- **[Roots](middleware/roots.md)** - Access control with URI templates
-- **[Elicitation](middleware/elicitation.md)** - Parameter review and collection
+- **[Onboarding New Servers](onboarding.md)** - Create your own wags server with middleware
+- **[Middleware Overview](middleware/overview.md)** - Understand available middleware features
diff --git a/mkdocs.yml b/mkdocs.yml
@@ -38,12 +38,13 @@ theme:
 nav:
   - Home: index.md
   - Quick Start: quickstart.md
+  - Onboarding Servers: onboarding.md
   - Middleware:
     - Overview: middleware/overview.md
     - TodoList: middleware/todo.md
     - Roots: middleware/roots.md
     - Elicitation: middleware/elicitation.md
-  - Benchmarks: benchmarks.md
+  - Running Evals: evals.md
 
 plugins:
   - search

diff --git a/servers/fastagent.config.yaml b/servers/fastagent.config.yaml
@@ -0,0 +1,15 @@
+mcp:
+  servers:
+    github:
+      transport: stdio
+      command: wags
+      args:
+        - start-server
+        - servers/github
+      env:
+        GITHUB_PERSONAL_ACCESS_TOKEN: ${GITHUB_PERSONAL_ACCESS_TOKEN}
+      roots:
+        - uri: https://github.com/anthropics/courses
+          name: "Anthropics Courses"
+        - uri: https://github.com/modelcontextprotocol/
+          name: "MCP Organization"