parcadei/Continuous-Claude-v3 / .claude/skills/mcp-chaining
mcp-chaining
VerifiedResearch-to-implement pipeline chaining 5 MCP tools with graceful degradation
MCP Chaining Pipeline
A research-to-implement pipeline that chains 5 MCP tools for end-to-end workflows.
When to Use
- Building multi-tool MCP pipelines
- Understanding how to chain MCP calls with graceful degradation
- Debugging MCP environment variable issues
- Learning the tool naming conventions for different MCP servers
What We Built
A pipeline that chains these tools:
| Step | Server | Tool ID | Purpose |
|---|---|---|---|
| 1 | nia | nia__search | Search library documentation |
| 2 | ast-grep | ast-grep__find_code | Find AST code patterns |
| 3 | morph | morph__warpgrep_codebase_search | Fast codebase search |
| 4 | qlty | qlty__qlty_check | Code quality validation |
| 5 | git | git__git_status | Git operations |
Key Files
scripts/research_implement_pipeline.py- Main pipeline implementationscripts/test_research_pipeline.py- Test harness with isolated sandboxworkspace/pipeline-test/sample_code.py- Test sample code
Usage Examples
# Dry-run pipeline (preview plan without changes)
uv run python -m runtime.harness scripts/research_implement_pipeline.py \
--topic "async error handling python" \
--target-dir "./workspace/pipeline-test" \
--dry-run --verbose
# Run tests
uv run python -m runtime.harness scripts/test_research_pipeline.py --test all
# View the pipeline script
cat scripts/research_implement_pipeline.py
Critical Fix: Environment Variables
The MCP SDK's get_default_environment() only includes basic vars (PATH, HOME, etc.), NOT os.environ. We fixed src/runtime/mcp_client.py to pass full environment:
# In _connect_stdio method:
full_env = {**os.environ, **(resolved_env or {})}
This ensures API keys from ~/.claude/.env reach subprocesses.
Graceful Degradation Pattern
Each tool is optional. If unavailable (disabled, no API key, etc.), the pipeline continues:
async def check_tool_available(tool_id: str) -> bool:
"""Check if an MCP tool is available."""
server_name = tool_id.split("__")[0]
server_config = manager._config.get_server(server_name)
if not server_config or server_config.disabled:
return False
return True
# In step function:
if not await check_tool_available("nia__search"):
return StepResult(status=StepStatus.SKIPPED, message="Nia not available")
Tool Name Reference
nia (Documentation Search)
nia__search - Universal documentation search
nia__nia_research - Research with sources
nia__nia_grep - Grep-style doc search
nia__nia_explore - Explore package structure
ast-grep (Structural Code Search)
ast-grep__find_code - Find code by AST pattern
ast-grep__find_code_by_rule - Find by YAML rule
ast-grep__scan_code - Scan with multiple patterns
morph (Fast Text Search + Edit)
morph__warpgrep_codebase_search - 20x faster grep
morph__edit_file - Smart file editing
qlty (Code Quality)
qlty__qlty_check - Run quality checks
qlty__qlty_fmt - Auto-format code
qlty__qlty_metrics - Get code metrics
qlty__smells - Detect code smells
git (Version Control)
git__git_status - Get repo status
git__git_diff - Show differences
git__git_log - View commit history
git__git_add - Stage files
Pipeline Architecture
+----------------+
| CLI Args |
| (topic, dir) |
+-------+--------+
|
+-------v--------+
| PipelineContext|
| (shared state) |
+-------+--------+
|
+-------+-------+-------+-------+-------+
| | | | | |
+---v---+---v---+---v---+---v---+---v---+
| nia |ast-grp| morph | qlty | git |
|search |pattern|search |check |status |
+---+---+---+---+---+---+---+---+---+---+
| | | | |
+-------v-------v-------v-------+
|
+-------v--------+
| StepResult[] |
| (aggregated) |
+----------------+
Error Handling
The pipeline captures errors without failing the entire run:
try:
result = await call_mcp_tool("nia__search", {"query": topic})
return StepResult(status=StepStatus.SUCCESS, data=result)
except Exception as e:
ctx.errors.append(f"nia: {e}")
return StepResult(status=StepStatus.FAILED, error=str(e))
Creating Your Own Pipeline
- Copy the pattern from
scripts/research_implement_pipeline.py - Define your steps as async functions
- Use
check_tool_available()for graceful degradation - Chain results through
PipelineContext - Aggregate with
print_summary()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |