Agent teams, swarms, memory and hooks for Claude Code and Codex, served over MCP.
Krusch Grounded Harness is installed from its publisher's own source and answers where it runs, so this marketplace is not in the path of a single call. There is no address here to send one to, and a panel that pretended otherwise would be showing you an answer we made up. Install it and call it from your own client — the Installation tab has the entry for each one.
What it does
Turn-0 workspace grounding, spin-loop blocker, and defect-only stop gate for AI coding agents.
Quickstart
# 1 — add it to Claude Code
claude mcp add --scope user krusch-grounded-harness-mcp '--' npx -y krusch-grounded-harness mcp
# 2 — ask your agent for something these tools do
# krusch_ground_workspace, krusch_resolve_scale_profile, krusch_grounded_session_start and 3 more
Krusch Grounded Harness costs nothing on mcprush: there is no plan to choose here, no cap to set and nothing mcprush can bill you for.
Collected from a public index. Nobody has claimed this account, so nothing here was written by its author — claim it if it is yours.
Where are you running it?
Every route below installs the same thing and ends at the same approval screen. This one runs on your machine: your client starts Krusch Grounded Harness as a process under your own user, with your files and your network, so the tool surface below is what it can do to your computer rather than to a server somewhere else. Its scan result is on the Security tab — read the surface before you approve it.
This is a public server: you run it yourself and this marketplace is not in the path. You need Node.js — the current LTS (24) is the safe choice; npx comes with it. Claude Code registers it in one command; --scope user makes it available in every project. The '--' is in quotes so that PowerShell passes it on.
claude mcp add --scope user krusch-grounded-harness-mcp '--' npx -y krusch-grounded-harness mcpReconnect, or start a new session, and the tools appear in the model’s tool list.
One config entry your client uses to start the process locally. A local server runs with your file system and your network, which is why it is priced without metering.
6 tools, with what each one reads, writes and reaches shown before you agree — the same list on every route above. Read the tool surface.
Tool surface
What the model actually sees. Descriptions are diffed on every release — see version history.
Read from its source code. A server that builds its tool list when it starts can list more tools than this.
Introspect local workspace schemas (CSV/TSV, JSON/JSONL, read-only SQLite DDL, source symbols, and 256/32 Raw-RAG docs) with 0 LLM calls, apply BM25 tool deferral, and emit a KV-cache-stable prompt bundle.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| goal | string | no | Task goal used to retrieve relevant doc chunks. |
| workspaceDir | string | no | Directory path to introspect. |
| model | string | no | Target model identifier for scale-aware directive. |
| activeParamsB | number | no | Optional active parameter count in billions. |
| totalParamsB | number | no | Optional total parameter count in billions. |
| thinkingEnabled | boolean | no | Whether native reasoning/thinking is enabled. |
| standingInvariants | array | no | Standing rules/invariants included in the static cache prefix. |
| tools | array | no | Full tool registry definitions to gate via BM25 deferral. |
| pinnedToolNames | array | no | Core tools kept eagerly exposed. |
| deferThreshold | integer | no | Tool count threshold triggering BM25 deferral. |
| maxGroundingTokens | integer | no | Maximum Turn-0 grounding token budget. |
Deterministically map a model identifier and parameter/thinking hints into SMALL_MODEL, MID_MODEL, or FRONTIER_REASONING harness policy.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| model | string | no | Model identifier (e.g., qwen2.5-coder:7b, gemma4:31b, gemini-4-argon). |
| activeParamsB | number | no | Optional active parameter count in billions. |
| totalParamsB | number | no | Optional total parameter count in billions. |
| thinkingEnabled | boolean | no | Whether native thinking mode is active. |
| thinkingLevel | string | no | Optional thinking level hint. |
| reasoningEffort | string | no | Optional reasoning effort hint. |
| explicitTier | string | no | Optional explicit tier override. |
| preferThinkingStepForSmallActive | boolean | no | Opt small-active MoE models into the compact PLAN->BIND->ACT directive. |
Create or reset a SQLite-tracked grounded session seeded with Turn-0 chunk IDs, file digests, scale profile, and schema token savings.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| session_id | string | no | Optional explicit session identifier. |
| goal | string | yes | User or task goal. |
| model | string | no | Target model identifier. |
| activeParamsB | number | no | — |
| totalParamsB | number | no | — |
| thinkingEnabled | boolean | no | — |
| anchor_date | string | no | Optional ISO anchor date (YYYY-MM-DD). |
| prefix_hash | string | no | Optional 64-char Turn-0 prompt cache prefix hash. |
| seen_chunk_ids | array | no | — |
| file_digests | object | no | — |
| all_tools | array | no | — |
| exposed_tools | array | no | — |
| turn0_tokens | integer | no | — |
| max_duplicate_writes | integer | no | — |
Check a proposed tool call against the frozen policy floor, secret-leak guard, duplicate-write guard, read memoization cache (maxReadStreak=0), and small-model placeholder soft-holds.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| session_id | string | yes | Active grounded session ID. |
| step | integer | no | 0-based or 1-based agent turn index. |
| tool_name | string | yes | Invoked tool name. |
| args | object | no | Tool call arguments. |
| mutating | boolean | no | Optional explicit mutability override. |
| suppress_turn0_reads | boolean | no | Short-circuit unchanged Turn-0 file reads. |
| block_spin_loop | boolean | no | Passively block identical repeat calls via BLOCK_SPIN_LOOP. |
Approve zero-overhead completion when pendingDiffsCount=0 and output has 0 objective defects while blocking unverified code diffs, empty/null/NaN bailouts, truncations, raw DataFrame/numpy reprs, and leaked <tool_call> tags.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| session_id | string | yes | Active grounded session ID. |
| assistant_text | string | no | Final assistant response text. |
| candidate_answer | no | Optional structured or scalar final answer value. | |
| finish_reason | string | no | Optional provider finish reason (e.g., STOP, MAX_TOKENS). |
| expect_non_empty_answer | boolean | no | Flag empty/null/NaN/[]/{} bailouts as output defects. |
| tool_calls | array | no | — |
| pending_diffs_count | integer | no | — |
| mutated_files_count | integer | no | — |
| latest_verification_passed | boolean | no | — |
| verification_command | string | no | — |
Retrieve session metrics including exploratory foraging ratio, memoized reads, spin loops blocked, output defects caught, partial-stdout timeouts, Turn-0 suppressed duplicates, schema token savings, and ordered events.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| session_id | string | yes | Grounded session ID. |
- Every tool, no call limit
- No card, and mcprush charges nothing for it
- Source published under MIT
- Runs on your machine — nothing of it reaches our gateway
- Nothing to cap, because mcprush bills nothing
What counts against your monthly calls
| Tool | Unit | Calls used | Out of the allowance |
|---|
Nothing here is billable: mcprush charges nothing to install Krusch Grounded Harness or to call it, at any volume.
Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.
| Term | Level | What it prices | Points |
|---|---|---|---|
| capability-exposure | high | capability blast radius (high) — client exposure if the model is manipulated | −6 |
| verification-discount | repo | publisher verification (public source) — no provenance, but the source is public and inspectable | −1 |
| coverage-honesty | source | inspection depth (source) — how much of the target the scan could see | −0 |
What the scan could actually read
A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.
No tools were enumerated, so prompt-injection, capability and toxic-flow analysis had no tool surface to inspect. To grade a package’s real runtime tools, scan the running server: --command "npx -y <package>".
Capability — what it could do if the model were manipulated
Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. high is the level these add up to.
| Tool | Capability tags | Why the tag was assigned |
|---|---|---|
| No per-tool rows for this release — see the note below. | ||
Toxic-flow graph
The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.
The public result for this release does not print the flow graph, so there is nothing to show here. That is not the same as "no paths were found": what the scan did read is above, under coverage.
Supply chain and provenance
This is the first scan of this surface here, so there is nothing yet to compare it against.
Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.
Release history
Always latest — the install command names no version, so it installs the newest release when it runs; the current release on file is 0.2.1. Pin a release below and the command asks for exactly that one.
No release note was published with this version.
One review per account, from a signed-in account that does not publish this listing. Nothing else is asked — you can say what you think before you install it. Publishers can reply once per review, and a review can be edited or deleted by whoever wrote it.
Any signed-in account that does not publish Krusch Grounded Harness can review it, once — before installing it or after. A review from an account that has installed it is marked as one.
Nobody has reviewed this listing. The rating on the card is the mean of the reviews written here and nothing else, so there is no rating until somebody writes the first — which takes a signed-in account that does not publish it, and nothing else.