agent-kernel-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DSH_REPO | No | Native helper script: absolute path to your deepseek-harness checkout (no default). | |
| AGENT_KERNEL_URL | Yes | Control-plane base URL, e.g. https://YOUR_WEB_HOST (not localhost when the API is remote). Required for the MCP server to start. | |
| AGENT_KERNEL_TOKEN | Yes | Session token (ak_session / Bearer). Required for API calls by the MCP server, although DSH overlays may allow an empty token at boot. | |
| AGENT_KERNEL_MCP_ROOT | No | Native DSH only: absolute path to this repo (set automatically by dsh/run-native-with-mcp.sh). |
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 |
|---|---|
| ak_healthB | Check agent-kernel API health. |
| ak_meB | Current authenticated user / setup gaps. |
| ak_list_projectsA | List catalog projects for the current user. |
| ak_get_projectB | Get one project by id. |
| ak_list_assignmentsA | List assignments for the current user. |
| ak_nudgeB | Nudge an assignment (kernel starts / continues an executor run). Prefer this over ad-hoc DSH autonomy when the control plane owns the schedule. |
| ak_scheduler_tickB | Run one agent-kernel scheduler tick (due assignments / nudges). |
| ak_attentionB | List observability attention items from agent-kernel. |
| ak_list_runsB | List recent runs. |
| ak_get_runA | Get one run by id. |
| ak_test_executorB | Live-test the configured executor (DSH) from the kernel. |
| ak_executor_settingsA | Read current user executor settings (secrets redacted by API). |
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 12 tools
Most tools clearly target distinct resources and actions, such as projects, assignments, runs, and health. The only mild ambiguity is between ak_nudge and ak_scheduler_tick, since both can trigger executor work, but their descriptions clarify the difference.
All tools share a consistent ak_ prefix and snake_case style, which helps readability. However, the set mixes verb_noun names like ak_list_projects with noun-only names like ak_health, ak_attention, and ak_me, so the pattern is not fully uniform.
12 tools is well within the ideal range and each tool serves a distinct operational area of the agent-kernel domain. The count feels appropriately scoped without redundancy or obvious bloat.
The surface covers listing and reading projects, assignments, and runs, plus scheduler and executor actions. Notable gaps include no way to update executor settings, cancel or retry runs, or get a single assignment by id, which limits full control-plane coverage.