@kud/mcp-harness-fme
This MCP server lets an MCP client inspect and manage Harness FME (Split.io) feature flags, segments, and related resources via a secure, confirm-gated toolset.
List workspaces, environments, traffic types, rollout statuses, feature flags, segments, and rule-based segments.
Create, update, archive, unarchive, delete, kill, and restore feature flags.
View and manage per-environment flag definitions/targeting rules, including adding segments to treatments.
Build Harness FME web-UI deep links for flags when optional account/org env vars are configured.
Manage classic segments: create, enable/disable, delete, list keys, and add/remove member keys.
Manage rule-based segments: create, update definitions, enable/disable per environment, and submit change requests with optional approvers.
Use pagination, filtering, and summary modes for large result sets.
Require
confirm: truefor destructive operations and supporttitle/commentfor workspaces that require them.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@@kud/mcp-harness-fme list all feature flags in workspace production"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for Harness FME (Split.io) — read and toggle feature flags.
Website · Documentation
Features
38 tools — covers workspaces, environments, feature flags, flag definitions, segments, rule-based segments, and change requests.
Kill & restore — instantly kill a flag to force all traffic to the default treatment, or restore it with a single tool call.
Safety guard — every destructive operation (delete, kill, archive, disable) requires
confirm: true, preventing accidental changes.Rule-based segments — create, update, enable, disable, and submit change requests for rule-based segments per environment.
Change request flow — submit segment definition changes with optional approvers for teams that require approval gates.
Zero-config startup — reads
MCP_HARNESS_FME_API_KEYfrom the environment and exits immediately if it is missing.
Related MCP server: Featureflow MCP Server
Install
Add to your MCP client config (see Usage below), or install globally to run manually:
npx --yes @kud/mcp-harness-fme@latestSet the environment variable MCP_HARNESS_FME_API_KEY to your Harness FME API key before starting the server.
Usage
This is a standard stdio MCP server — it works with any MCP client (Claude Desktop, Claude Code, Cursor, Windsurf, Cline, Zed, …). Add it to your client's MCP config:
{
"mcpServers": {
"harness-fme": {
"command": "npx",
"args": ["--yes", "@kud/mcp-harness-fme@latest"],
"env": {
"MCP_HARNESS_FME_API_KEY": "your_api_key"
}
}
}
}Most clients read this mcpServers shape — Claude Desktop's config file, Cursor's .cursor/mcp.json, Windsurf, Cline, and so on. For Claude Code, there's a CLI shortcut:
claude mcp add --transport stdio --scope user harness-fme \
--env MCP_HARNESS_FME_API_KEY=your_api_key \
-- npx --yes @kud/mcp-harness-fme@latestTo enable the get_flag_url deep-link tool, also set two optional keys — add them to the env block above (or as extra --env flags for the CLI). See Configuration for where to find their values.
"env": {
"MCP_HARNESS_FME_API_KEY": "your_api_key",
"MCP_HARNESS_FME_ACCOUNT_ID": "your_account_id",
"MCP_HARNESS_FME_ORG_GUID": "your_org_guid"
}Leave them out and every other tool still works — get_flag_url just reports what's missing.
Available tools
Tool | Description |
| List all FME workspaces in the account |
| List all environments in a workspace |
| List all traffic types in a workspace |
| List rollout status definitions for a workspace |
| List feature flags in a workspace (filter by tag, status, name) |
| Get metadata for a specific feature flag |
| Create a new feature flag for a given traffic type |
| Update a flag's description, tags, owners, or rollout status |
| Permanently delete a feature flag ( |
| Archive a flag, removing it from active use ( |
| Restore a previously archived feature flag |
| Kill a flag in an environment — forces default treatment ( |
| Restore a killed feature flag in an environment |
| List flag targeting rules in an environment |
| Get treatments and targeting rules for a flag in an environment |
| Activate a flag in an environment with treatments and targeting rules — pass |
| Fully replace a flag's targeting rules in an environment — pass |
| Remove a flag's targeting rules from an environment ( |
| Add a segment to a flag treatment via safe read-modify-write (idempotent; avoids full-replace) — pass |
| Build a Harness FME web-UI deep-link for a flag — pass workspace/flag/environment by name or id (needs two |
| List all segments in a workspace (max 50 per page) |
| List classic segments activated in a specific environment |
| List a classic segment's member keys in an environment — check whether a key is already in the segment |
| Create a new classic segment in a workspace under a traffic type |
| Activate a classic segment in an environment |
| Add member keys to a classic segment in an environment — pass |
| Remove specific member keys from a classic segment ( |
| Remove a classic segment from an environment ( |
| Permanently delete a classic segment from a workspace ( |
| List all rule-based segments in a workspace |
| Get a rule-based segment's workspace-level metadata |
| Create a new rule-based segment in a workspace |
| Permanently delete a rule-based segment ( |
| List rule-based segment definitions in an environment |
| Update a rule-based segment's rules in an environment |
| Activate a rule-based segment in an environment |
| Remove a rule-based segment from an environment ( |
| Submit a change request for a segment definition with optional approval flow |
Configuration
The server reads these environment variables:
Variable | Required? | Used by |
| Required | Every tool — server exits at startup if absent |
| Alternative | Read first when set, so one exported key also serves |
| Optional |
|
| Optional |
|
The account ID and org GUID are Harness platform identifiers that the API does not expose, so get_flag_url reads them from env. Grab them once from any flag's URL in the Harness FME web UI:
https://app.harness.io/ng/account/<ACCOUNT_ID>/all/fme/orgs/<ORG_SLUG>/projects/<PROJECT>/org/<ORG_GUID>/ws/<WORKSPACE_ID>/splits/<FLAG_ID>/env/<ENV_ID>/definitionURL segment | Where it goes |
| → |
| → |
| Resolved by |
| Resolved by |
⚠ The URL has two org-ish segments:
/orgs/(plural) is the human slug;/org/(singular) is the GUID you want forMCP_HARNESS_FME_ORG_GUID. Grab the singular one.
With those two set, get_flag_url needs only a workspace, flag, and environment (name or id) — it resolves everything else. Without them, it returns a message telling you what to set, and every other tool works normally.
Development
git clone https://github.com/kud/mcp-harness-fme.git
cd mcp-harness-fme
npm install
npm run dev # run from source with tsx
npm run inspect:dev # MCP Inspector at http://localhost:5173
npm test # vitest
npm run build # compile to dist/Environment variables are documented in Configuration above. Tools are registered in src/index.ts; their handlers live in src/tools.ts and call @kud/harness-fme, the client library this server shares with the terminal tool @kud/harness-fme-cli.
📚 Full documentation → mcp-harness-fme/docs
Available Tools
30 toolsadd_segment_to_treatmentA
Add a segment to a specific treatment of a feature flag via read-modify-write — safer than update_flag_definition's full replace, where a dropped field silently removes existing targeting. Idempotent: a no-op if the segment is already on the treatment. For workspaces where list_workspaces reports requiresTitleAndComments: true, pass title (and comment).
Inputs: segment (required), flag_name (required), treatment (required), workspace_id (required), +3 more.
archive_feature_flagA
Archive a feature flag — removes it from active use while preserving history. Subject to OPA policy checks (409 on failure).
Inputs: flag_name (required), workspace_id (required), confirm.
create_feature_flagC
Create a new feature flag for a given traffic type
Inputs: name (required), traffic_type (required), workspace_id (required), description.
create_flag_definitionA
Create (activate) a feature flag definition in a specific environment with treatments and targeting rules. For workspaces where list_workspaces reports requiresTitleAndComments: true, pass title (and comment) or the API rejects the write with a 400.
Inputs: flag_name (required), definition (required), workspace_id (required), environment_id (required), +2 more.
create_rule_based_segmentB
Create a new rule-based segment in a workspace under a specific traffic type
Inputs: name (required), traffic_type (required), workspace_id (required), description.
create_rule_based_segment_change_requestA
Submit a change request for a rule-based segment definition — supports approval flow via approvers
Inputs: title (required), workspace_id (required), environment_id (required), operation_type (required), +3 more.
delete_feature_flagB
Permanently delete a feature flag from a workspace — irreversible
Inputs: flag_name (required), workspace_id (required), confirm.
delete_flag_definitionA
Remove a feature flag definition from an environment — the flag itself remains, but loses its targeting rules in that environment
Inputs: flag_name (required), workspace_id (required), environment_id (required), confirm.
delete_rule_based_segmentA
Permanently delete a rule-based segment from a workspace — environment-level configs must be removed separately
Inputs: segment_name (required), workspace_id (required), confirm.
disable_rule_based_segment_definitionA
Disable (remove) a rule-based segment from a specific environment — workspace-level metadata is preserved
Inputs: segment_name (required), workspace_id (required), environment_id (required), confirm.
enable_rule_based_segment_definitionB
Enable (activate) a rule-based segment in a specific environment — creates an empty definition that can then be configured via update
Inputs: segment_name (required), workspace_id (required), environment_id (required).
get_feature_flagA
Get metadata for a specific feature flag
Inputs: flag_name (required), workspace_id (required).
get_flag_definitionA
Get the targeting rules and treatment definition of a feature flag in a specific environment
Inputs: flag_name (required), workspace_id (required), environment_id (required).
get_flag_urlA
Build a deep-link to a feature flag's definition in the Harness FME web UI. Pass workspace, flag, and environment by name or id — the org slug, project, flag id, and env id are resolved for you. Requires MCP_HARNESS_FME_ACCOUNT_ID and MCP_HARNESS_FME_ORG_GUID env vars on the server (those IDs are not exposed by the API — copy them once from a flag URL in the browser).
Inputs: flag (required), workspace (required), environment (required).
get_rule_based_segmentB
Get a rule-based segment's workspace-level metadata by name
Inputs: segment_name (required), workspace_id (required).
kill_feature_flagB
Kill (disable) a feature flag in an environment — forces all traffic to the default treatment
Inputs: flag_name (required), workspace_id (required), environment_id (required), confirm.
list_environmentsA
List all environments in a workspace
Inputs: workspace_id (required).
list_feature_flagsC
List feature flags in a workspace
Inputs: workspace_id (required), tag, name, limit, +2 more.
list_flag_definitionsA
List all feature flag definitions (targeting rules) in a specific environment. Full definitions are large; pass summary: true to get name/id only when you just need to locate a flag.
Inputs: workspace_id (required), environment_id (required), limit, offset, +1 more.
list_rollout_statusesA
List rollout status definitions for a workspace (e.g. Killed, Permanent, Ramping) — use the returned IDs to filter feature flags by rollout status
Inputs: workspace_id (required).
list_rule_based_segment_definitionsB
List rule-based segment definitions in a specific environment
Inputs: workspace_id (required), environment_id (required).
list_rule_based_segmentsB
List all rule-based segments in a workspace
Inputs: workspace_id (required), limit, offset.
list_segmentsC
List all segments in a workspace
Inputs: workspace_id (required), limit, offset.
list_traffic_typesB
List all traffic types in a workspace
Inputs: workspace_id (required).
list_workspacesA
List all FME workspaces in the account
Inputs: limit, offset.
restore_feature_flagA
Restore (re-enable) a killed feature flag in an environment
Inputs: flag_name (required), workspace_id (required), environment_id (required).
unarchive_feature_flagC
Unarchive a previously archived feature flag, restoring it to active use
Inputs: flag_name (required), workspace_id (required).
update_feature_flagC
Update a feature flag's description, tags, or owners (partial update)
Inputs: flag_name (required), workspace_id (required), tags, owners, +2 more.
update_flag_definitionA
Fully replace a feature flag definition (treatments, targeting rules) in an environment. For workspaces where list_workspaces reports requiresTitleAndComments: true, pass title (and comment) or the API rejects the write with a 400.
Inputs: flag_name (required), definition (required), workspace_id (required), environment_id (required), +2 more.
update_rule_based_segment_definitionC
Update a rule-based segment definition in an environment (rules, exclusions, matchers)
Inputs: definition (required), segment_name (required), workspace_id (required), environment_id (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 30 tools
Most tools have distinct purposes, but potential overlap exists between add_segment_to_treatment and update_flag_definition, and among the various segment list tools (list_segments, list_rule_based_segments, list_rule_based_segment_definitions). Descriptions help clarify, but an agent might still misselect.
All tool names follow a consistent verb_noun pattern in snake_case, e.g., create_feature_flag, list_environments, update_flag_definition. No mixing of conventions.
With 30 tools, the set is slightly over the typical well-scoped range (3-15), but it covers a complex domain (feature flag management with many sub-resources). Each tool seems justified, though some granularity could be reduced.
The tool set provides comprehensive CRUD and lifecycle operations for feature flags, segments, environments, workspaces, and more. Includes advanced operations like rollback, change requests, and deep-links. No obvious gaps for the intended domain.
Maintenance
Related MCP Connectors
Users, organizations, roles, permissions, applications and feature flags.
Feature flags, A/B experiments, kill switches, dynamic config & i18n — Shipeasy MCP server.
Manage ConfigDirector feature flags, remote configs, and targeting rules from AI assistants.
Feature flag service for safe rollouts across the Hive agent fleet
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables interaction with LaunchDarkly's feature flag platform through AI clients. Supports managing feature flags, AI configs, and their variations with operations like create, update, delete, and targeting configuration.131,828 npm27MIT
- AlicenseAqualityDmaintenanceEnables AI assistants to manage Featureflow feature flags, including creating and updating features, controlling feature states across environments, and managing projects, environments, and targeting rules through natural language.226 npmMIT

ConfigCat MCP Serverofficial
AlicenseCqualityBmaintenanceProvides access to ConfigCat's management API for feature flag and configuration management, enabling CRUD operations on entities like feature flags, configs, environments, and products, as well as SDK documentation.961,606 npm17MIT
Featureflipofficial
AlicenseAqualityBmaintenanceManage Featureflip feature flags, targeting and rollouts from AI agents and editors. 19 tools over the public Management API let agents read flag state, create and toggle flags, edit targeting rules, manage segments and find stale flags from Claude Code, Cursor, Cline or any MCP-compatible client. Every tool is annotated read-only or destructive. Apache-2.0.43984 npmApache 2.0