ollama-mcp-server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
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 | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| run_ollama_taskB | Delegates a heavy code, text generation, or analysis sub-task to a local or cloud Ollama model. |
| list_ollama_modelsA | Lists Ollama models currently available (pulled locally, or signed-in cloud models) so a suitable one can be picked for run_ollama_task. |
| summarize_outputA | Condenses large text (logs, command output, file dumps) into a short summary using a local Ollama model, so the full text never has to enter the caller's own context. Pass |
| outline_fileA | Returns the indexed declaration outline for one supported source file. Use this before read_symbol to choose focused source ranges. |
| read_symbolA | Returns one indexed symbol's source range, body, and one line of surrounding context. Obtain the stable symbol ID from outline_file. |
| find_symbolA | Finds indexed declarations by stable ID, exact name or qualified name, or partial name. Use the returned stable IDs with the relationship tools. |
| find_referencesA | Finds indexed references that resolved to one declaration. Unresolved occurrences are deliberately excluded. |
| find_callersA | Finds indexed call sites whose callee resolved to one declaration. Each result includes its resolution quality. |
| find_calleesA | Finds indexed call sites made from within one declaration. Each result includes its resolution quality. |
| hybrid_retrieveB | Ranks source symbols plus structural, documentation, package-script, and Git-commit evidence without embeddings by default. Returned evidence is source-backed context, not a verified answer. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 10 tools
The set splits into two clearly distinguishable clusters: model delegation (run_ollama_task, list_ollama_models, summarize_output) and code-index lookup (outline_file, read_symbol, find_symbol, find_references, find_callers, find_callees, hybrid_retrieve). Within the symbol cluster, find_references and find_callers are adjacent concepts, but descriptions clarify that one covers resolved references and the other covers call sites only. summarize_output is essentially a specialisation of run_ollama_task, so mild overlap remains, but the other boundaries are clear.
Nine of the ten tools use snake_case with an action prefix (run_/list_/summarize_/outline_/read_/find_) and a concrete object. hybrid_retrieve breaks the pattern, because its prefix is an adjective rather than a verb, so it is the one moderate deviation. The naming is otherwise predictable and readable.
Ten tools is well within the ideal range and each tool covers a distinct operation across the two capabilities. Nothing looks like junk or a duplicate entry, and the server does not overwhelm the agent with tool choice.
The symbol-navigation surface (outline, read, find by name, references, callers, callees, hybrid retrieval) is unusually complete for a code-indexing capability. The main gaps are an explicit tool to list or read files from a repository and an index-status or refresh tool; the LLM-delegation side also lacks a model-pull or conversation-continuation operation, but agents can work around these gaps.