Qualithm Operator MCP server
Commit Nov 15, 2026Lets an AI client provision and manage the Qualithm platform, including teams, devices, credentials, and automations.
Also known as qualithm/operator-mcp
- Stars
- 0
- Tools
- 11
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 15, 2026
- License
- Apache-2.0
What is Qualithm Operator MCP?
Qualithm Operator lets an AI client provision and manage the Qualithm platform, including teams, devices, credentials, and automations. It exposes 11 tools, such as code, message, action.
Summary generated from the project's README
About Qualithm Operator MCP server
Provisioning and managing the Qualithm platform through an AI client is what operator-mcp enables for developers and platform engineers. This Go-based server exposes the Qualithm management API as a set of tools over stdio, allowing agents to handle resources like teams, devices, credentials, and automations. It uses the same underlying client as the official CLI to ensure that human and agent actions remain consistent across the platform.
To set up the server, you must provide a member API token starting with the qmt_ prefix via the QUALITHM_API_TOKEN environment variable or the --token flag. You can also specify a custom management API URL. The server provides a uniform result envelope for every tool, which includes fields like ok, code, and message, making it easy for agents to handle errors or plan changes using the dryRun feature. While it manages most platform resources, sensitive actions like billing tier changes and session mutations are restricted to human users.
What you can do with it
- 1Plan platform changes safelyAsk the AI to simulate a device update or team creation without actually applying the changes.
- 2Manage device credentialsInstruct an agent to rotate or mint new credentials for specific devices on the platform.
- 3Automate platform workflowsHave an AI client create, trigger, or monitor the status of platform automations.
- 4Monitor system observabilityRequest the AI to retrieve telemetry data or audit logs to troubleshoot platform activity.
Written from the project's README. Every use case is checked against a quote from it.
Qualithm Operator MCP tools 11
Detected from the READMEcodemessageactiondataauthnot_foundconflictrate_limitedunavailableapierror
Configuration
QUALITHM_API_TOKEN
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablequalithm/operator-mcp is public and reachablegithub.com/qualithm/operator-mcp ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-13github.com/qualithm/operator-mcp ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model11 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNo tool named like delete, drop or execPass
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubOperator MCP
Go MCP server for managing teams, scopes and devices on Qualithm. It exposes the platform
management API — teams, members, scopes, devices, credentials, enrollments, authorities, API tokens,
automations, dashboards, observability, and read-only workspace and billing state — as MCP tools
over stdio, backed by the same operator client as the qualithm CLI, so both return the same
results.
Features
- Provisioning tools — one MCP tool per verb over every resource listed under Tools.
- Uniform result envelope — every tool returns the same
{ ok, code, message, retryAfterSeconds, dryRun, action, data }shape, so agents branch on one structure regardless of tool. - Per-call dry-run — mutating tools accept
dryRun; when set, the change is planned and the intended request is reported without being sent. - Stable error codes — failures carry a code mirroring the CLI's exit-code contract (
auth,not_found,conflict,rate_limited,unavailable,api,error). - Bearer auth — authenticates with a member API token (prefix
qmt_).
Installation
go install github.com/qualithm/operator-mcp/cmd/qualithm-mcp@latest
A container image is published to GHCR on each release:
docker pull ghcr.io/qualithm/operator-mcp:latest
Quick Start
The server authenticates with a member API token (prefix qmt_). Provide it via --token or the
QUALITHM_API_TOKEN environment variable; point at an environment with --url or
QUALITHM_API_URL (defaults to https://api.qualithm.com). It speaks the Model Context Protocol
over stdio.
export QUALITHM_API_TOKEN=qmt_...
qualithm-mcp
Register it with an MCP-capable agent, for example:
{
"mcpServers": {
"qualithm-operator": {
"command": "qualithm-mcp",
"env": {
"QUALITHM_API_TOKEN": "qmt_..."
}
}
}
}
| Flag | Env | Description |
|---|---|---|
--url |
QUALITHM_API_URL |
management API base URL |
--token |
QUALITHM_API_TOKEN |
member API token (qmt_…) |
--team |
QUALITHM_TEAM |
team to act on (defaults to the token's) |
--version |
— | print version and exit |
--help |
— | print usage and exit |
Tools
| Resource | Tools |
|---|---|
| authorities | list_authorities · create_authority · revoke_authority |
| enrollments | list_enrollments · create_enrollment · revoke_enrollment |
| credentials | list_credentials · mint_credential · issue_cert · rotate_credential · revoke_credential |
| devices | list_devices · get_device · create_device · update_device · delete_device |
| device commands | list_device_commands · send_device_command · get_device_capabilities |
| scopes | list_scopes · get_scope · create_scope · update_scope · delete_scope · list_scope_devices |
| teams | list_teams · create_team · get_team · update_team · delete_team · get_team_device_state |
| members & invites | list_team_members · add_team_member · get_team_member · update_team_member · remove_team_member · list_team_invites · create_team_invite · revoke_team_invite |
| automations | list_automations · create_automation · get_automation · update_automation · delete_automation · run_automation · list_automation_runs · list_automation_templates · trigger_automation · create_automation_trigger_secret |
| dashboards | list_dashboards · create_dashboard · get_dashboard · update_dashboard · delete_dashboard |
| observability | get_telemetry · stream_events · get_usage · get_audit_log |
| account | get_account · list_capabilities · list_roles · list_sessions · get_session · get_communication_preferences · list_zones |
| api tokens | list_api_tokens · revoke_api_token |
| billing (read-only) | get_billing_summary · list_invoices · preview_tier_change |
Money-moving billing routes (tier changes, add-ons, checkout and portal sessions) and
account/session mutations stay human-only. The full route-to-tool mapping, including every excluded
route's rationale, is cmd/coverage-check/coverage.json — CI fails when a new platform route ships
without a tool or a recorded rationale.
Result envelope
Every tool returns the same structured payload:
| Field | Meaning |
|---|---|
ok |
whether the call succeeded (a dry-run counts as success) |
code |
failure classification when ok is false |
message |
human-readable error message when ok is false |
retryAfterSeconds |
seconds to wait before retrying an unavailable failure |
dryRun |
true when a mutation was planned but not applied |
action |
the planned request (method, path) for dry-run results |
data |
the resource payload returned by the API on success |
Error codes
| Code | Meaning |
|---|---|
auth |
401 / 403 |
not_found |
404 |
conflict |
409 |
rate_limited |
429 |
unavailable |
503 |
api |
other non-2xx |
error |
transport / unexpected |
unavailable means a backend was briefly unavailable: retry the same call after
retryAfterSeconds.
A paused zone also returns auth (403) with message Zone rejects creation in this environment;
treat it as "the zone is closed here" and pick an open zone. Production opens de-fra-a and
sg-sin-a; lower environments scope to sg-sin-a.
Development
Prerequisites
- Go 1.26+
Setup
make install-tools
This installs golangci-lint, goimports, govulncheck and gosec into $GOPATH/bin (~/go/bin
by default). Put that directory on your PATH.
Building
make build
Testing
make test # unit tests with race detector
make test-coverage # with coverage report
Agent eval
cmd/agent-eval is the end-to-end agent-native eval: it drives the documented agent path against a
live environment — create an enrollment via the MCP tools, claim the device, connect to the gateway,
publish telemetry, and read it back — and records every step where the agent had to supply
information the surfaces did not provide (the friction log). The Agent Eval workflow runs it
weekly against sandbox and files a deduped issue on failure.
go build -o /tmp/qualithm-mcp ./cmd/qualithm-mcp
go run ./cmd/agent-eval -server-bin /tmp/qualithm-mcp -token qmt_...
Linting & Formatting
make lint
make fmt
make vet
Security Tooling
make audit # govulncheck
make gosec # standalone gosec scan
make lint # golangci-lint (includes gosec checks via .golangci.yaml)
.github/workflows/audit.yaml runs govulncheck and gosec daily.
Minimum Supported Go Version
Go 1.26+.
License
Apache-2.0
Qualithm Operator MCP questions
What tools does Qualithm Operator MCP provide?
It exposes 11 tools, including code, message, action, data, auth.
How does the server authenticate?
The server uses Bearer authentication with a member API token that must start with the prefix 'qmt_'.
Can I test changes without applying them?
Yes, mutating tools accept a dryRun parameter which allows you to plan a change and see the intended request without sending it.
What happens if a tool fails?
Failures return stable error codes such as auth, not_found, conflict, rate_limited, unavailable, api, or error.