Atlas / MCP servers / harness / harness-mcp-v2

harness-mcp-v2BLOCK

Name: harness-mcp-v2
Author: harness

mcp/harness/mcp-server

MCP server by harness · harness · mcp · mcp server · Audited 2026-10-12

⚑ Report / re-scan Source ↗

Harness’ official MCP Server

Verdict

BLOCK

Grade

Trust score

69 /100

Exposed tools

200 259r · 21w · 2d

Transport

stdio · streamable-http

License

MIT

Stars

105

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Harness MCP Server 2.0

[](https://mcptoplist.com/server/glama%2Fharness%2Fmcp-server)

An MCP (Model Context Protocol) server that gives AI agents full access to the Harness.io platform through 11 consolidated tools and 261 resource types.

Why Use This MCP Server

Most MCP servers map one tool per API endpoint. For a platform as broad as Harness, that means 240+ tools — and LLMs get worse at tool selection as the count grows. Context windows fill up with schemas, and every new endpoint means new code.

This server is built differently:

11 tools, 261 resource types. A registry-based dispatch system routes harness_list, harness_get, harness_create, etc. to any Harness resource — pipelines, services, environments, orgs, projects, feature flags, cost data, and more. The LLM picks from 11 tools instead of hundreds.
Full platform coverage. 42 default toolsets spanning CI/CD, GitOps, Feature Flags, Cloud Cost Management, Security Testing, Chaos Engineering, Database DevOps, Internal Developer Portal, Software Supply Chain, Infrastructure as Code Management, Release Management, Governance, Service Overrides, Knowledge Graph, and more. Opt-in Ansible and observability-evaluation coverage is available when needed.
Multi-project workflows out of the box. Agents discover organizations and projects dynamically — no hardcoded env vars needed. Ask "show failed executions across all projects" and the agent can navigate the full account hierarchy.
35 prompt templates. Pre-built prompts for common workflows: build & deploy apps end-to-end, debug failed pipelines, review DORA metrics, triage vulnerabilities, optimize cloud costs, audit access control, plan feature flag rollouts, review pull requests, approve pending pipelines, and more.
Works everywhere. Stdio transport for local clients (Claude Desktop, Cursor, Devin Desktop), HTTP transport for remote/shared depl

Read from source at commit f5ffcf9c8f3dOBSERVED · 2026-10-12

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add harness-mcp-v2 --env GITHUB_TOKEN=${GITHUB_TOKEN} --env HARNESS_API_KEY=${HARNESS_API_KEY} -- npx -y [email protected]

claude-desktop

{
  "mcpServers": {
    "harness-mcp-v2": {
      "command": "npx",
      "args": [
        "-y",
        "[email protected]"
      ],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}",
        "HARNESS_API_KEY": "${HARNESS_API_KEY}"
      }
    }
  }
}

Exposed tools (200)

259 read · 21 write · 2 destructive. Blast radius: 2 tools can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.

Tool	Risk	Description
`Acme`	read
`Deploy`	write	Prod deploy
`Service`	read	A service
`action`	read	Filter audit logs by action type
`activity_type`	read	Comma-separated activity types
`after`	read	Only entries created at/after this timestamp (unix millis)
`agent_id`	read	Scope-prefixed GitOps agent identifier (required).
`all`	read	When true, include delegates in child org/project scopes under the requested scope. Omit for the current scope only.
`annotations`	read	Filter by annotation key(s). Only types with at least one matching annotation are returned.
`annotator_type`	read	human \| automated
`anomaly_view`	read	View type for anomaly grouping
`app_name`	read	GitOps application name
`applied_at_end`	read	APPLIED-tab only: appliedAtEndTime epoch ms. Ignored for OPEN.
`applied_at_start`	write	APPLIED-tab only: appliedAtStartTime epoch ms. Ignored for OPEN. Do not send the Open-tab calendar as days_back.
`approval_status`	read	Approval status filter
`approval_type`	read	Approval type filter
`artifact_id`	read	Artifact ID to list components for
`artifact_type`	read	Filter by artifact type (e.g., CONTAINER, FILE)
`aspect`	read	Which BA data to fetch
`audit_resource_type`	read	Filter audit logs by resource type (renamed from resource_type to avoid conflict with MCP parameter)
`authentication`	read	Provider authentication object {type, spec}. Use secret references only (e.g. spec.tokenRef=
`auto_upgrade`	read	Filter by auto upgrade status
`before`	read	Only entries created before this timestamp (unix millis)
`branch`	read	Branch to filter executions
`budget_sort_type`	read	Sort field
`categories`	read	Security violation categories to include
`category`	read	Connector category filter
`checkout-experiment`	read	d
`cloud_filter`	read	Filter by cloud provider
`commander`	read	Filter by incident commander (multi-value). Validated against the project
`committer`	read	Filter by committer
`comparisons`	read	Optional filter to a subset of the experiment
`connectivity_statuses`	read	Filter by connectivity status
`connector_connectivity_modes`	read	Filter by connectivity mode
`connector_identifiers`	read	Filter by connector identifiers (comma-separated)
`connector_names`	read	Filter by connector names (comma-separated)
`cost_buckets`	read	Cost bucket(s) within the cost category. Comma-separated for multiple (e.g.
`cost_category`	read	Cost category name to filter by (must pair with cost_buckets). Use with cost_recommendation_filter to list teams/buckets.
`created_after`	read	Only alerts created at or after this time (ISO-8601, e.g. 2026-05-01T00:00:00Z)
`created_before`	read	Only alerts created at or before this time (ISO-8601)
`created_date_from`	read	Filter evaluations created after this epoch-millis timestamp. ⚠️ IMPORTANT: when omitted, the upstream policy-mgmt API silently defaults to
`created_date_to`	read	Filter evaluations created before this epoch-millis timestamp
`cursor`	read	Cursor for the next page (from previous response pagination.next_cursor)
`dataset_id`	read	Parent dataset UUID
`date_end`	read	End date for metric calculation
`date_start`	write	Start date for metric calculation
`days_back`	read	OPEN/IGNORED freshness window: lastProcessedAt within this many UTC days (CCM default 4). Do NOT map the UI date picker here — that picker is appliedAt for APPLIED recs. Omitted automatically when recommendation_states is APPLIED only.
`dbinstance_id`	read	Instance identifier — required
`dbschema_id`	read	Schema identifier — required to list instances
`delegate_group_identifier`	read	Filter by delegate group identifier
`delegate_instance_filter`	read	Filter by delegate instance
`delegate_name`	read	Filter delegates by name
`delegate_type`	read	Delegate type filter
`deleted`	read	Include soft-deleted packages
`dependency_type`	read	Filter by dependency type (DIRECT or TRANSITIVE)
`deplay-action-01`	read	optional desc
`deployed_after`	read	Only deploys at or after this time (ISO-8601, e.g. 2026-05-01T00:00:00Z)
`deployed_before`	read	Only deploys at or before this time (ISO-8601)
`deployment_type`	read	Filter by deployment type (e.g. Kubernetes, ECS)
`description`	read	Filter by connector description
`drift_id`	read	Drift ID from scs_sbom_drift calculate response
`end_time`	read	End time in ISO 8601 format (e.g. 2025-07-10T23:59:59Z). Default: now.
`enforcement_id`	read	Enforcement ID (from artifact overview violations.enforcementId)
`entity`	read	Filter by entity identifier
`entity_identifier`	read	Entity identifier in
`entity_type`	read	Filter global templates by entity type
`env_type`	read	Environment type filter
`environment`	read	Filter by environment (multi-value, OR-combined). Validated against the project
`environment_id`	read	FME environment ID (get from fme_environment)
`epss`	read	EPSS probability bucket (single select).
`epss_percentile`	read	EPSS percentile bucket (single select).
`eval_id`	read	Evaluation UUID
`event_groups`	read	Event groups to include (multi-value, OR-combined). Omit for all named groups — which is not every event:
`event_type_ids`	read	Filter by base event type IDs
`exception_id`	read	Filter by exception ID
`exclude_rego`	read	Exclude rego source from list response
`exclude_repo_patterns`	read	Glob patterns matching target.name on repository targets (sto-core excludeRepoPatterns).
`execution_id`	read	Filter by pipeline plan execution ID (e.g.
`exemption_statuses`	read	Comma-separated statuses
`experiment_id`	read	Experiment ID (get from fme_experiment). Required.
`exploitability`	read	Exploitability filter.
`favorites`	read	Only return entities the current user has favorited
`feature_flag_name`	read	Feature flag name whose definitions to list. Required.
`feature_type`	read	Productivity feature type
`filter_type`	read	Which groups to include at this scope
`folder_id`	read	Filter by CCM folder ID
`folder_name_pattern`	read	Filter folders by name pattern (substring match)
`freeze_status`	read	Filter by freeze status
`git_ref`	read	Git reference (branch/tag) filter
`global`	read	When true, accesses global templates (list: passes isGlobal=true; get: forces accountIdentifier to __GLOBAL_TEMPLATES_ACCOUNT_ID__)
`granularity`	read	Time granularity
`group`	read	Kubernetes API group filter (e.g.
`group_by`	read	Group results by field. Use predefined fields (region, product, etc.) OR any label key name (env, team, app, environment, etc.)
`harness_create`	write
`harness_delete`	destructive
`harness_describe`	read
`harness_diagnose`	read
`harness_execute`	write
`harness_get`	read
`harness_list`	read
`harness_schema`	read
`harness_search`	read
`harness_status`	read
`harness_update`	write
`has_module`	read	Filter by module presence (hasModule)
`host_name`	read	Filter by host name
`identifier_filter`	read	Filter by policy identifier
`identifiers`	read	Optional list of File Store identifiers to fetch details for (repeatable query param)
`ids`	read	Filter to specific metric IDs
`impacted_service`	read	Filter by impacted Harness service identifier (multi-value, OR-combined). Validated against the project
`import_request_id`	write	Required. importRequestId from the import execute response.
`include_child_scopes`	read	Include evaluations from child scopes
`include_directories`	read	If true, include directories in the listing as well as files.
`include_meta`	read	Include list-envelope counts (
`include_parent_scopes`	read	Include settings that exist only at parent scopes
`include_policy_set_count`	write	Include count of policy sets referencing each policy
`inherited`	read	Include rules inherited from parent spaces (true/false)
`inheriting_credentials_from_delegate`	read	Filter connectors inheriting credentials from delegate
`integration_type`	read	AI coding assistant type
`is_cluster_hourly_data`	read	Whether to use cluster hourly data
`is_cluster_query`	read	Whether this is a cluster query
`is_custom`	write	(get only) Whether the check is a custom check. Set to true for custom checks; defaults to false.
`issue_types`	read	Comma-separated issue types
`key`	read	Metadata key whose values to enumerate (required)
`kind`	read	Kubernetes resource kind filter
`kinds`	read	Filter by object kind(s). Pass one or more OBJECT_KIND_* values.
`label`	read	Filter by label
`lifecycle`	read	Comma-separated list of lifecycles to filter on (e.g.
`limit`	read	Result limit
`match_type`	read	How name is matched. Ignored when name is omitted.
`metadata`	read	Filter by metadata using key:value format
`metadata_only`	read	When true, fetches only template metadata via list-metadata — faster than full list
`metric`	read	DORA metric to fetch
`metric_ids`	read	Optional display filter to a subset of metrics. Never affects multiple comparison correction, which is always computed from the metric
`metric_set_id`	write	Filter by metric set UUID
`metric_type`	read	Metric to retrieve
`migration_type`	read	Filter by migration type
`min_amount`	read	Minimum actual amount threshold
`min_anomalous_spend`	read	Minimum anomalous spend threshold
`min_saving`	read	Minimum monthly savings. Default 1 (Open/Ignored tabs). Default 0 when recommendation_states is APPLIED only. Pass 0 to include sub-$1 recs.
`module`	read	Filter audit logs by module
`module_type`	read	Module type (e.g. CD, CI)
`my-name`	read	desc
`my_deployments`	read	Show only my deployments
`name`	read	Filter delegate tokens by name
`namespace`	read	Kubernetes namespace filter
`new-apm-probedvdv`	read
`new-custom-script-action-8s3`	read	optional desc
`node_execution_id`	read	Node execution ID to filter by step
`object_kind`	read	Filter by type kind
`object_type`	read	Type of object to list (e.g. Table) — required for list
`offset`	read	Pagination offset
`only_favorites`	read	Only favorited projects
`only_true_positive_issue_types`	read	Issue types that must have a TRUE_POSITIVE triage verdict to stay in scope
`order`	write	Sort order
`order_by_direction`	write	Order direction
`order_by_field`	write	Field to order by
`org_id`	write	Optional — not supported; use fme_segment_definition execute actions for Harness-native segment key operations.
`origin_statuses`	read	Comma-separated origin statuses.
`origins`	read	Comma-separated origin layers.
`oss_risk_filter`	read	Filter by OSS risk category. Comma-separated values from: DEFINITE_EOL, DERIVED_EOL, CLOSE_TO_EOL, UNMAINTAINED, OUTDATED
`owned_by_me`	read	Only return entities owned by the current user
`owner`	read	Comma-separated list of owner references to filter on
`package_id`	read	Package ID to scope the file search
`package_ids`	read	Comma-separated package IDs
`package_kind`	read	Package kind (huggingface model/dataset, terraform module/provider)
`package_name`	read	Filter by package name
`package_type`	read	Filter registries by package type
`page`	read	Page number (0-based)
`page_token`	read	Pagination token from a previous list response
`parent_name`	read	Filter to this parent name (Feature Flag/AI Config/Config name matching parent_type). Unknown or mismatched name returns 404.
`parent_type`	read	Parent kind to list. Required. FEATURE_FLAG and AI_CONFIG are implemented; CONFIG returns 404.
`path`	read	File path filter
`perspective_id`	read	Cost perspective identifier
`perspective_name`	read	Filter by perspective name(s) (comma-separated)
`phase_id`	write	Optional authored phase id. When set, limits the list to the latest artifacts_reported attempt of that phase.
`phase_identifier`	read	Filter to activities in a specific phase
`pipeline_id`	read	Pipeline identifier to filter executions
`pipeline_ids`	read	Comma-separated pipeline IDs
`pod-delete`	destructive	Pod delete chaos
`policy_set_ref`	write	OPA policy set reference
`principal_type`	read	Principal type filter
`priority`	read	Filter by priority option id (multi-value, OR-combined)
`priyanshu-http-probe-1`	read	ops desc
`product_names`	read	Comma-separated scanner product names (e.g.
`profile_id`	read	Business alignment profile ID
`project_id`	write	Optional — not supported; use fme_segment_definition execute actions for Harness-native segment key operations.
`provider`	read	LLM provider (required)
`purl`	read	Package URL of the component (e.g. pkg:npm/[email protected]) — required
`query`	read	The semantic search query (e.g.
`reachability`	read	Reachability filter.
`recommendation_states`	read	OPEN (default, matches Open tab), APPLIED, or IGNORED (expands to IGNORED+TEMPORARILY_IGNORED). Comma-separated.
`region`	read	Optional provider region (e.g. for Bedrock/Vertex)
`registry_id`	read	Registry ID to scope the file search
`registry_ids`	read	Comma-separated registry IDs to filter within
`repo_creds_id`	read	Filter by repository credentials ID
`reported_after`	read	Only incidents reported after this time (ISO-8601)
`reported_before`	read	Only incidents reported before this time (ISO-8601)
`resource_group_identifier`	read	Resource group identifier filter
`resource_group_identifiers`	read	Filter by resource group identifiers (comma-separated)

Trust audit

BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	FAIL
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (7 observation(s))
Network: declared (10 observation(s))
Shell: declared (6 observation(s))
Dependencies: not all pinned
Secrets in source: found

Findings (25)

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95

src/audit/sinks/otel.ts:13

return await (Function("m", "return import(m)")(moduleName) as Promise<unknown>);

Why it matters. evaluates text as code

Fix. remove; use a parser or a dispatch table

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

Dockerfile:71

CMD ["node", "-e", "const port=process.env.PORT||'3000';fetch(`http://127.0.0.1:${port}/health`).then(r=>{if(!r.ok)process.exit(1)}).catch(()=>process.exit(1))"]

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

src/registry/toolsets/gitops.ts:1689

"   {list:{elements:[{cluster:'staging', url:'https://1.2.3.4'}, {cluster:'prod', url:'https://2.3.4.5'}]}}\n" +

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

src/registry/toolsets/gitops.ts:1704

"   {matrix:{generators:[{list:{elements:[{cluster:'staging', url:'https://1.2.3.4'}]}}, {git:{repoURL:'...', revision:'HEAD', directories:[{path:'apps/*'}]}}]}}\n" +

MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321

tests/client/harness-client.test.ts:380

let token = "keycloak-access-token-1";

MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321

tests/client/harness-client.test.ts:384

token = "keycloak-access-token-2";

MEDIUMFilesystem / path · mcp.destructive_tools · CWE-22, CWE-59

harness_delete, pod-delete

Why it matters. 2 tool(s) can delete or overwrite

Fix. prefer a read-only mode or scoped tokens; the page states the blast radius

LOWInventory / provenance · inv.hidden_file · CWE-1104

.mcpbignore

.mcpbignore

Why it matters. hidden member outside the usual dotfiles

Fix. review its purpose

LOWFilesystem / path · fs.credential_store · CWE-22, CWE-59

tests/registry/chaos-probe.test.ts:322

env: [{ name: "KUBECONFIG", value: "/root/.kube/config" }],

Why it matters. touches a credential store

LOWFilesystem / path · fs.credential_store · CWE-22, CWE-59

tests/registry/chaos-probe.test.ts:336

env: [{ name: "KUBECONFIG", value: "/root/.kube/config" }],

Why it matters. touches a credential store