chaining-mcp-server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MCP_SERVERS | No | JSON string containing additional MCP server configurations | |
| GITHUB_TOKEN | No | GitHub Personal Access Token required for awesome-copilot tools | |
| AWESOME_COPILOT_ENABLED | No | Set to 'false' to disable awesome-copilot integration | |
| DISABLE_THOUGHT_LOGGING | No | Set to 'true' to disable sequential thinking thought logging | |
| SEQUENTIAL_THINKING_AVAILABLE | No | Set to 'true' to enable sequential thinking integration | |
| RELIABILITY_MONITORING_ENABLED | No | Set to 'true' to enable reliability monitoring |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_mcp_serversA | Lists all discovered MCP servers on the system |
| analyze_toolsC | Analyzes available tools from discovered MCP servers |
| generate_route_suggestionsC | Generates optimal route suggestions for a given task |
| analyze_with_sequential_thinkingC | Analyzes complex workflows using sequential thinking |
| get_tool_chain_analysisC | Gets comprehensive analysis of available tools and suggested routes |
| sequentialthinkingA | Think one step with the Mitosis agent: your thought is recorded as an observation in AgentState, Needle decides the next action (call_tool/revise/complete/escalate), and the result feeds the next step. Pass sessionId to continue a session. |
| search_instructionsC | Searches custom instructions based on keywords in their descriptions |
| load_instructionC | Loads a custom instruction from the repository |
| brainstormingA | Generate creative ideas via a generative model (requires CHAINING_LLM_ENABLED with OPENROUTER_API_KEY; fails honestly without a key — no template ideas) |
| workflow_orchestratorC | Execute complex multi-server workflows across the MCP ecosystem with dependency management and error handling |
| get_current_timeA | Get current time in a specific timezone |
| convert_timeC | Convert time between timezones |
| get_promptB | Get a specific prebuilt prompt by ID |
| search_promptsC | Search for prompts by keywords, category, or tags |
| get_resource_setB | Get a specific resource set by ID |
| search_resource_setsC | Search for resource sets by keywords, category, or tags |
| validate_tool_chainA | Validate tool chains for correctness, dependencies, and potential issues. Checks for circular dependencies, tool availability, and parameter compatibility |
| analyze_tool_chain_performanceC | Analyze performance metrics and efficiency of tool chains. Provides execution time estimates, complexity analysis, and optimization suggestions |
| workflow_statusA | Get the current status and step results of a workflow executed by workflow_orchestrator or agent_run |
| workflow_cancelB | Request cancellation of a running workflow |
| list_skillsA | List all discovered agent skills (name, description, files) from the local skills catalog. Read-only; never executes skill scripts. |
| search_skillsA | Search the skills catalog by keywords against names and descriptions. Returns ranked matches for chaining. |
| get_skillA | Load a skill's full instructions (SKILL.md body) plus its file manifest, for injection into agent context or harness use. |
| suggest_skill_chainB | Plan a task with Needle over registry tools and attach skill recommendations per step (deterministic catalog match, labeled). Combines skills+tools into one executable chain. |
| llm_queryA | Execute a direct query using the internal LLM engine (OpenRouter/OpenAI compatible, disabled by default) |
| llm_decompose_taskB | Decompose a complex goal into ordered subtasks with recommended tool categories. [Legacy compat API — routes via the Needle planner when MITOSIS_AGENT_ENABLED=true.] |
| llm_suggest_routeA | Use LLM intelligence to score and rank optimal multi-tool execution routes. [Legacy compat API — prefer agent_run when MITOSIS_AGENT_ENABLED=true.] |
| llm_summarizeB | Compress verbose tool outputs, logs, or multi-step execution results |
| agent_runA | Run a task through the Needle agent runtime: plans with the local model, executes tools via the workflow orchestrator, escalates to OpenRouter only on low confidence or failure. Requires MITOSIS_AGENT_ENABLED=true and a fetched engine (npm run needle:fetch). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| MCP Servers | Returns a JSON list of all discovered MCP servers |
| Available Tools | Returns a JSON list of all available tools from discovered servers |
| Tool Chain Analysis | Returns a JSON summary of the current analysis state |
| Prebuilt Prompts | Returns a JSON collection of all available prebuilt prompts for common development tasks |
| Resource Sets | Returns a JSON collection of curated resource sets for different development scenarios |
| Prompts Overview | Returns a JSON overview of available prompts by category and complexity level |
| Awesome Copilot Collections | Returns a JSON collection of all available awesome-copilot collections with their metadata |
| Awesome Copilot Instructions | Returns a JSON collection of all available awesome-copilot instructions with their metadata |
| Awesome Copilot Integration Status | Returns a JSON object with the current status of awesome-copilot integration |
| Sequential Thinking State | Returns a JSON object with the current state of sequential thinking sessions, including thought history and active session status |
| Workflow Orchestrator Status | Returns a JSON object with the status of active and completed workflow orchestrations, including execution progress and results |
| Tool Chaining Resources | Returns a JSON collection of comprehensive tool chaining resources including prompts and resource sets specifically designed for complex development workflows and orchestration patterns |
| Tool Chaining Overview | Returns a JSON overview of available tool chaining resources organized by category and complexity level, providing insights into the tool chaining capabilities |
| Server Health Status | Returns real-time health status of all tools, discovery services, and integrations |
| Cache Statistics | Returns cache hit/miss statistics and discovery cache state |
| Internal LLM Engine Status | Returns status and configuration of internal LLM engine (never exposes keys) |
| Internal LLM Usage Statistics | Returns token usage and call volume for the internal LLM engine |
| Needle Agent Runtime Status | Returns configuration and readiness of the local Needle agent runtime and escalation policy (never exposes keys) |
TDQS
Scored across 29 tools
Several tools occupy the same planning/analysis space: llm_suggest_route, generate_route_suggestions, get_tool_chain_analysis, analyze_tools, analyze_with_sequential_thinking, and suggest_skill_chain all produce route/task plans with overlapping descriptions. sequentialthinking and analyze_with_sequential_thinking are also easy to confuse, and legacy llm_* tools duplicate newer agent_run functionality. Only the skill/prompt/resource get/search tools are clearly separable.
Most tools use verb_noun snake_case, but there are notable deviations: sequentialthinking and brainstorming have no separators, workflow_status/workflow_orchestrator are noun-first, workflow_cancel reverses the expected verb_noun order, and agent_run is noun+verb. The llm_ prefix is applied inconsistently, with brainstorming and llm_query both invoking LLM capabilities but only one being prefixed.
At 29 tools the surface is overgrown for a chaining server. Several tools are explicitly legacy compat APIs that duplicate newer entry points, and unrelated utilities like get_current_time and convert_time add noise. A focused chaining server could express the same capabilities in roughly half the tools.
The chaining lifecycle is well covered: discover servers and skills, load skills and instructions, plan/validate/analyze chains, execute via workflow_orchestrator or agent_run, monitor status, and cancel. Minor gaps exist, such as no direct ability to list running workflows or execute a skill standalone, but agents can work around these. The legacy/duplicate APIs do not create serious dead ends.