Skip to content

Commit 9776f01

Browse files
committed
Updated README and CLAUDE
1 parent 62a67ad commit 9776f01

File tree

2 files changed

+3
-34
lines changed

2 files changed

+3
-34
lines changed

CLAUDE.md

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
# Load claude permanent context
22

3+
@README.md
34
@.claude/context/overview.md
45
@.claude/context/architecture.md
5-
@.cladue/context/commands.md
6+
@.claude/context/commands.md

README.md

Lines changed: 1 addition & 33 deletions
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@ Too many MCP servers work great when you're testing them locally, but then fail
1010

1111
## The Two-Step Verification Process
1212

13-
### 1. 🏥 `doctor` - Automatic Health Check
13+
### 1. 🏥 `doctor` - Spec compliance check
1414

1515
Runs diagnostics on your MCP server without needing any configuration or knowledge of your server's functionality. Catches spec violations that cause LLMs to fail or behave unpredictably.
1616

@@ -148,38 +148,6 @@ evals:
148148
mcp-server-tester test tests.yaml --server-config server-config.json
149149
```
150150

151-
## Why Both Doctor and Test?
152-
153-
Think of it like shipping software:
154-
155-
- **Doctor** = Linter/compiler checks (catches structural problems)
156-
- **Test** = Unit/integration tests (verifies business logic)
157-
158-
You need both! A server might pass all functional tests but still fail with LLMs due to spec violations that `doctor` would catch.
159-
160-
## Common Issues Doctor Catches
161-
162-
1. **Incorrect error codes** - LLMs expect specific JSON-RPC error codes
163-
2. **Missing error messages** - LLMs need descriptive errors to retry/recover
164-
3. **Protocol violations** - Subtle spec issues that confuse LLM tool parsers
165-
4. **Missing annotations** - Hints that help LLMs use tools correctly
166-
167-
## Best Practices for LLM-Friendly Servers
168-
169-
1. **Run `doctor` first** - Fix spec compliance before testing functionality
170-
2. **Test with real prompts** - Use eval tests with prompts users would actually write
171-
3. **Handle errors gracefully** - LLMs rely on clear error messages
172-
4. **Use tool annotations** - Add hints like `readOnlyHint` and `openWorldHint`
173-
5. **Test edge cases** - LLMs will find creative ways to use your tools
174-
175-
## Full Documentation
176-
177-
- [Writing Tests](./docs/writing-tests.md)
178-
- [Server Configuration](./docs/server-config.md)
179-
- [Doctor Diagnostics](./docs/doctor-diagnostics.md)
180-
- [Eval Testing Guide](./docs/eval-testing.md)
181-
- [Examples](./examples/)
182-
183151
## Contributing
184152

185153
We welcome contributions! Please ensure all PRs:

0 commit comments

Comments
 (0)