Skip to main content
Glama

Container Host AIops

Audit availability: Real writes require a committed pending audit row and stop before contacting the platform if storage is unavailable. Reads and previews may continue with a warning; new audit rows store outcome summaries rather than full responses.

Disclaimer: Community-maintained open-source project. Not affiliated with, endorsed by, or sponsored by Docker, Inc., Portainer.io, or any container-platform vendor. "Docker", "Portainer" and all product/trademark names belong to their respective owners. MIT licensed.

Governed AI-ops for non-orchestrator container hosts — the Docker Engine API (over a local unix socket or a TCP host), Portainer (its management API, which also proxies Docker), and Podman (a rootful/rootless service socket speaking the Docker-compatible API plus libpod-native endpoints) — with a built-in governance harness: a unified audit log over both MCP and CLI, a runaway/budget safety guard, and undo-token recording. It records every operation; it does not decide whether a write is permitted — that is the agent's or the account's call. Multi-platform by construction: a registry keyed by platform means a per-target platform field (docker / portainer / podman) selects the API shape, and another host family could be added later without touching the ops/CLI/MCP layers. Exercised against a live Docker Engine 27.5.1 daemon (doctor, overview, the three flagship analyses, and a governed stop_container with audit + undo recorded); the Portainer and Podman API paths are covered by the mock suite only.

What it does

Three flagship signature analyses, plus the guarded reads and writes around them:

  • Restart-loop RCA — inspect containers for restart count + exit code, flag the crash-looping ones (restartCount over threshold, or restarting/dead, or a non-zero exit), and map each to a likely cause + action from the exit code (137 OOM/SIGKILL, 143 SIGTERM, 139 segfault, 127 bad entrypoint, …), with a tail of logs. Every ranking carries its numbers, not a black-box verdict.

  • Resource-pressure analysis — a one-shot CPU%/memory% sample per running container vs its configured limits, flagging each "near" (≥ 80% of a threshold) or "over", with a recommendation (raise a limit, set a missing memory limit, scale out).

  • Image & volume bloat — dangling images + dangling volumes + build cache from system/df, totalled into prune candidates with reclaimable bytes.

Related MCP server: observability-aiops

What works

  • CLI (container-host-aiops ...): init, overview, container, image, volume, network, system, stack, pod, analyze, manage, secret, doctor, mcp.

  • MCP server (container-host-aiops mcp or container-host-aiops-mcp): 38 tools (29 read, 9 write), every one wrapped with the bundled @governed_tool harness.

  • Connection layer: Docker over a unix socket (httpx.HTTPTransport(uds=...)) or a TCP host; Portainer over HTTPS with an X-API-Key token that also proxies the Docker API of a managed endpoint; Podman over its rootful/rootless service socket (autodetected: $XDG_RUNTIME_DIR/podman/podman.sock first, then /run/podman/podman.sock) speaking the Docker-compat layer (paths reused wholesale) plus libpod-native endpoints. A local Docker/Podman socket needs no secret — the socket's file permissions are the boundary.

  • Encrypted credentials: the Portainer API token lives in an encrypted store ~/.container-host-aiops/secrets.enc (Fernet + scrypt) — never plaintext on disk. Unlock with a master password from CONTAINER_HOST_AIOPS_MASTER_PASSWORD (MCP/CI) or an interactive prompt (CLI).

  • Reversibility: mutating writes fetch the real before-state first and record a faithful inverse (stop↔start; update_container restores prior CPU/memory limits). Irreversible ops (remove_container, prune_images, prune_volumes, recreate_stack) capture the before-state for audit but declare no undo.

  • Safety: every state-changing CLI op supports --dry-run and requires double confirmation; every write MCP tool takes a dry_run preview — and prune previews list what would be removed + reclaimable bytes before doing it.

What this tool does, and does not, decide

It delivers container-host operations — reads and writes — accurately and efficiently, and records every one of them. It does not decide whether a write is allowed to happen. That is the agent's judgement, or the permission of the account you connect it with: point it at a Docker socket mounted read-only, or a Portainer account without write scope, and the writes fail at the server — the place that actually owns the permission.

So there is no read-only switch, no policy file, no approval gate to configure. The one thing the tool guarantees is that nothing is silent: every call, over MCP and over the CLI alike, lands an audit row in ~/.container-host-aiops/audit.db, and destructive writes still capture their before-state and record an inverse where one exists.

Each tool declares a risk_level, kept in agreement with its [READ]/[WRITE] documentation tag by a test, and carried into the audit row as a descriptive tier — so a reviewer can see at a glance that a row was a high-risk delete. It is a label, not a gate.

Capability matrix (38 MCP tools)

Domain

Tools

Count

R/W

Overview

overview

1

read

Containers

list_containers, inspect_container, container_logs, container_stats, container_top, container_restart_summary

6

read

Images

list_images, inspect_image, dangling_images, image_disk_usage

4

read

Volumes

list_volumes, inspect_volume, dangling_volumes

3

read

Networks

list_networks, inspect_network

2

read

System

system_info, system_version, system_df, system_events

4

read

Stacks

list_endpoints, list_stacks, stack_detail (Portainer), list_compose_stacks (docker+podman)

4

read

Pods (Podman)

list_pods

1

read

Analyses (flagship)

restart_loop_rca, resource_pressure_analysis, image_and_volume_bloat

3

read

Writes

remove_container, prune_images, prune_volumes, recreate_stack

4

write (high)

restart_container, stop_container, start_container, update_container

4

write (medium)

The three analyses accept injected data for offline analysis, or pull live from a configured target. list_endpoints/list_stacks/stack_detail require a portainer target; list_compose_stacks (Compose project rollup by the com.docker.compose.project label, with per-stack health) works on docker or podman; list_pods requires a podman target (Docker/Portainer have no pod concept).

Platform support matrix

Capability

docker

portainer

podman

Container / image / volume / network / system reads

✅

✅ (proxied)

✅ (compat)

Flagship analyses (restart-loop RCA, resource pressure, bloat)

✅

✅

✅ (compat)

Lifecycle + prune writes (stop/start/restart/remove/update/prune)

✅

✅

✅ (compat)

Compose-stack rollup (list_compose_stacks)

✅

✅ (proxied)

✅

Portainer endpoints / stacks / recreate_stack

—

✅

—

Podman pods (list_pods, libpod)

—

—

✅

Quick start

As a Claude Code plugin

One install gives an agent both the skill and the MCP server:

/plugin marketplace add AIops-tools/marketplace
/plugin install container-host-aiops@aiops-tools

The MCP server is fetched with uv and pinned to the package version this plugin declares, so an audit row can be traced back to the code that wrote it. Credentials are still configured with container-host-aiops init — see below.

As an OpenClaw plugin

The same bundle is published on ClawHub, where one install delivers the skill and its MCP server together:

openclaw plugins install clawhub:@zw008/container-host-aiops
openclaw skills info container-host-aiops          # expect: Visible to model: yes

Restart the OpenClaw gateway afterwards so it loads the plugin. The MCP server is fetched with uv, pinned to this exact release, so uvx has to be on PATH — without it the skill still installs but reports Visible to model: no. Credentials are configured exactly as below.

As a CLI or standalone MCP server

uv tool install container-host-aiops          # or: pipx install container-host-aiops
container-host-aiops init                      # wizard: add a Docker/Podman socket or Portainer target
container-host-aiops doctor                    # verify config, secrets, connectivity
container-host-aiops overview                  # one-shot host health
container-host-aiops analyze restart-loop      # crash-looping containers + cause/action
container-host-aiops container list --running  # running containers

Run as an MCP server (stdio):

export CONTAINER_HOST_AIOPS_MASTER_PASSWORD=...   # only needed for Portainer targets
container-host-aiops-mcp

Where that password then lives: an exported variable is readable by every process this shell starts and is recorded by shell history. On a shared or long-lived host, prefer the interactive prompt, or inject it from a secret manager for the life of the one command that needs it.

Governance

Every operation — MCP and CLI — passes through the bundled @governed_tool harness. It records; it does not authorize (see above).

  • Audit — every call (params, outcome summary, status, duration, risk tier, and any operator-supplied approver/rationale) is logged to ~/.container-host-aiops/audit.db (relocatable via CONTAINER_HOST_AIOPS_HOME). The CLI writes the same row the MCP path does — there is no unaudited entry point.

  • Runaway guard — a safety backstop, not an authorization gate: the same call hammered in a tight loop trips a circuit breaker so a stuck agent can't burn unbounded calls/time. Disable with CONTAINER_HOST_RUNAWAY_MAX=0; optional hard ceilings via CONTAINER_HOST_MAX_TOOL_CALLS / CONTAINER_HOST_MAX_TOOL_SECONDS.

  • Undo recording — reversible writes record an inverse descriptor built from the fetched before-state.

  • Risk tier — a descriptive label on the audit row derived from risk_level; it gates nothing.

Scope

This is the container-host member of the AIops-tools family (governed AI-ops with audit + budget + undo), for single-host Docker / Portainer / Podman. It is deliberately NOT for a cluster orchestrator, a hypervisor, a storage appliance, a backup product, or OT / industrial edge — those are separate tools/lines.

Missing a capability?

Coverage is intentionally a curated subset of the Docker Engine + Portainer + Podman (libpod) APIs. Missing a call, or want another container host family? Open an issue or PR — contributions welcome.

Verification status

  • Docker — exercised against a live Docker Engine 27.5.1 daemon: doctor and overview connected over the local socket, the restart-loop and image/volume bloat analyses were run against real crash-looping containers and real reclaimable data, and a governed stop_container wrote a row to the audit DB and recorded a working undo descriptor.

  • Portainer and Podman — mock-validated only. Those API paths are modelled from each project's public API shape and have not been exercised against a live server.

  • The full checklist — what the mock suite guarantees, what the Docker run already satisfied, and what a Portainer/Podman run still has to prove — is in docs/VERIFICATION.md. container-host-aiops doctor is the fastest live check on any platform.

Available Tools

38 tools
container_logsA
Read-only

[READ] Tail the last N log lines of a container (stdout + stderr).

Args: container_id: Container id or name. tail: Number of lines from the end (1..2000, default 100). target: Target name from config; omit for the default.

Returns an envelope: {"lines": [...], "returned": N, "limit": L, "truncated": bool}. When "truncated" is true the container has older log history than was returned — re-run with a higher tail rather than treating this as the container's complete log.

Inputs: container_id (required), tail, target.

container_restart_summaryA
Read-only

[READ] Restart-count + exit-code summary across containers, worst-first.

Args: all_states: True (default) inspects all containers; False only running. target: Target name from config; omit for the default.

Inputs: target, all_states.

container_statsA
Read-only

[READ] One-shot CPU%/memory% snapshot for a container (stream=false).

Args: container_id: Container id or name. target: Target name from config; omit for the default.

Inputs: container_id (required), target.

container_topB
Read-only

[READ] Processes running inside a container (like docker top).

Args: container_id: Container id or name. target: Target name from config; omit for the default.

Inputs: container_id (required), target.

dangling_imagesA
Read-only

[READ] Untagged (dangling) images + reclaimable bytes — prune candidates.

Args: target: Target name from config; omit for the default.

Inputs: target.

dangling_volumesA
Read-only

[READ] Dangling volumes (unreferenced) + reclaimable bytes — prune candidates.

Args: target: Target name from config; omit for the default.

Inputs: target.

image_and_volume_bloatA
Read-only

[READ] Total dangling images + volumes + build cache into prune candidates.

Sums dangling images, dangling volumes, and build cache (from system/df) into reclaimable-byte prune candidates, largest first. Pass the three read payloads for pure analysis, or a target to pull live.

Args: dangling_images: Injected {danglingCount, reclaimableBytes, ...}; skips live. dangling_volumes: Injected {danglingCount, reclaimableBytes, ...}; skips live. df: Injected system/df summary {buildCache:{count, totalBytes}, ...}. target: Target name from config; omit for the default.

Inputs: df, target, dangling_images, dangling_volumes.

image_disk_usageA
Read-only

[READ] Image disk usage from system/df (total, shared, reclaimable).

Args: target: Target name from config; omit for the default.

Inputs: target.

inspect_containerA
Read-only

[READ] Full inspect of one container (config, state, mounts, network).

Args: container_id: Container id or name. target: Target name from config; omit for the default.

Inputs: container_id (required), target.

inspect_imageB
Read-only

[READ] Inspect an image plus its build history (layers, sizes, commands).

Args: image_id: Image id or name:tag. target: Target name from config; omit for the default.

Inputs: image_id (required), target.

inspect_networkA
Read-only

[READ] Inspect one network (driver, IPAM subnet/gateway, attached containers).

Args: network_id: Network id or name. target: Target name from config; omit for the default.

Inputs: network_id (required), target.

inspect_volumeA
Read-only

[READ] Inspect one named volume (driver, mountpoint, options, usage).

Args: name: Volume name. target: Target name from config; omit for the default.

Inputs: name (required), target.

list_compose_stacksA
Read-only

[READ] Group containers into Compose projects with a per-stack health rollup.

Reconstructs docker compose / podman compose projects from the 'com.docker.compose.project' label over the Docker-compat layer, so it works on a docker OR podman target (no Portainer needed). Each stack rolls up to healthy (all running) / degraded (some) / down (none).

Args: target: Target name from config; omit for the default.

Inputs: target.

list_containersA
Read-only

[READ] List containers, bucketed by state, with compact rows.

Args: all_states: True (default) lists all containers; False only running. target: Target name from config; omit for the default.

Inputs: target, all_states.

list_endpointsA
Read-only

[READ] Portainer endpoints (managed hosts): id, name, type, status, url.

Requires a portainer target.

Args: target: Target name from config; omit for the default.

Inputs: target.

list_imagesB
Read-only

[READ] List images (tags, size, dangling), largest first.

Args: all_images: True includes intermediate layers; default only top-level. target: Target name from config; omit for the default.

Inputs: target, all_images.

list_networksA
Read-only

[READ] List Docker networks, bucketed by driver.

Args: target: Target name from config; omit for the default.

Inputs: target.

list_podsA
Read-only

[READ] List Podman pods (Podman-only): id, name, status, member containers.

A pod groups containers sharing namespaces — a Podman-native concept with no Docker equivalent, read over the libpod API. Requires a podman target; on a docker/portainer target it returns a teaching error (pods do not exist there).

Args: target: Target name from config; omit for the default.

Inputs: target.

list_stacksA
Read-only

[READ] Portainer stacks (Compose/Swarm): id, name, type, endpoint, status.

Requires a portainer target.

Args: target: Target name from config; omit for the default.

Inputs: target.

list_volumesB
Read-only

[READ] List named volumes (name, driver, mountpoint, scope).

Args: target: Target name from config; omit for the default.

Inputs: target.

overviewA
Read-only

[READ] One-shot host overview: version + container state rollup + disk.

Call this first to triage a container host before drilling into a specific container, image, or volume.

Args: target: Target name from config; omit for the default.

Inputs: target.

prune_imagesA
Destructive

[WRITE][risk=high] Prune images (dangling by default). No undo.

dry_run LISTS the images that would be removed and the reclaimable bytes before doing anything.

Args: dangling_only: True (default) prunes only untagged images; False also prunes images unused by any container. dry_run: If True, list what would be removed + reclaimable bytes. target: Target name from config; omit for the default.

Inputs: target, dry_run, dangling_only.

prune_volumesA
Destructive

[WRITE][risk=high] Prune unreferenced volumes. No undo.

Docker removes only ANONYMOUS unused volumes by default (its behaviour since 23.0); named unused volumes survive unless all_unused is set. The dry_run preview is scoped by the same flag, so it describes the call you are about to make — and when scoped to the default it still reports the named unused volumes it will not touch (alsoUnusedNamed*) rather than hiding that space.

Args: all_unused: Also remove NAMED unused volumes (docker volume prune -a). Irreversible, and named volumes usually hold data someone meant to keep. dry_run: If True, list what would be removed + reclaimable bytes. target: Target name from config; omit for the default.

Inputs: target, dry_run, all_unused.

recreate_stackA
Destructive

[WRITE][risk=high] Redeploy (recreate) a Portainer stack. No undo.

Captures the stack definition for audit before Portainer recreates it; there is no clean inverse for a redeploy. Requires a portainer target. Pass dry_run=True to preview.

Args: stack_id: Portainer stack id. endpoint_id: Endpoint id the stack runs on; omit to use the stack's own. dry_run: If True, preview without recreating. target: Target name from config; omit for the default.

Inputs: stack_id (required), target, dry_run, endpoint_id.

remove_containerA
Destructive

[WRITE][risk=high] Remove a container (captures full inspect first). No undo.

The complete inspect JSON is captured before deletion for the audit trail; there is no clean inverse for a removed container. Pass dry_run=True to preview.

Args: container_id: Container id or name. force: Force-remove a running container (SIGKILL). remove_volumes: Also remove anonymous volumes attached to it. dry_run: If True, preview without removing. target: Target name from config; omit for the default.

Inputs: container_id (required), force, target, dry_run, +1 more.

resource_pressure_analysisA
Read-only

[READ] Rank containers by CPU/memory pressure vs their limits + recommend.

Pulls a one-shot CPU%/mem% sample for each running container (or uses injected 'samples'), flags each 'near' (>= 80% of a threshold) or 'over' (>= threshold), and attaches a recommendation (raise a limit, set a missing memory limit, scale out). Ranks worst-first by the higher of CPU%/mem%. Every row carries its numbers.

Args: cpu_threshold: CPU% at/above which a container is over pressure (default 80). mem_threshold: Memory% at/above which a container is over pressure (default 80). samples: Injected rows {id, name, cpuPercent, memPercent, memUsageBytes, memLimitBytes}; skips live collection. target: Target name from config; omit for the default.

Inputs: target, samples, cpu_threshold, mem_threshold.

restart_containerA
Destructive

[WRITE][risk=medium] Restart a container (captures prior state for audit).

A restart has no meaningful inverse, so no undo is recorded. Pass dry_run=True to preview.

Args: container_id: Container id or name. timeout: Seconds to wait for graceful stop before killing (default 10). dry_run: If True, preview without restarting. target: Target name from config; omit for the default.

Inputs: container_id (required), target, dry_run, timeout.

restart_loop_rcaA
Read-only

[READ] Find crash-looping containers and map each to a cause + action.

The flagship restart RCA: inspects containers for restart count + exit code, flags the crash-looping ones (restartCount >= threshold, or restarting/dead, or a non-zero exit), attaches a likely cause + recommended action from the exit code (137 OOM/SIGKILL, 143 SIGTERM, 139 segfault, 127 bad entrypoint, …), and a tail of logs. Every ranking carries its numbers. Pass 'containers' for pure analysis, or a target to pull live.

Args: restart_threshold: Restart count at/above which a container is looping (default 3). containers: Injected rows {id, name, state, restartCount, exitCode, oomKilled, error}; skips live collection. logs_by_id: Optional {containerId: [logLine, ...]} log tails to attach. target: Target name from config; omit for the default.

Inputs: target, containers, logs_by_id, restart_threshold.

stack_detailA
Read-only

[READ] One Portainer stack in detail (env, entrypoint, resource control).

Requires a portainer target.

Args: stack_id: Portainer stack id. target: Target name from config; omit for the default.

Inputs: stack_id (required), target.

start_containerA
Destructive

[WRITE][risk=medium] Start a stopped container. Inverse: stop it.

Pass dry_run=True to preview.

Args: container_id: Container id or name. dry_run: If True, preview without starting. target: Target name from config; omit for the default.

Inputs: container_id (required), target, dry_run.

stop_containerA
Destructive

[WRITE][risk=medium] Stop a running container. Inverse: start it.

Captures whether it was running so the harness records a start undo. Pass dry_run=True to preview.

Args: container_id: Container id or name. timeout: Seconds to wait for graceful stop before killing (default 10). dry_run: If True, preview without stopping. target: Target name from config; omit for the default.

Inputs: container_id (required), target, dry_run, timeout.

system_dfB
Read-only

[READ] Disk-usage breakdown: images, containers, volumes, build cache.

Args: target: Target name from config; omit for the default.

Inputs: target.

system_eventsA
Read-only

[READ] Recent daemon events over the last N seconds, rolled up by type+action.

Args: since: Look-back window in seconds (1..86400, default 3600). event_type: Filter to one event type (container/image/volume/network); omit for all. target: Target name from config; omit for the default.

Returns an envelope: {"events": [...], "returned": N, "limit": L, "truncated": bool, "total": T}. When "truncated" is true only the most recent events fit — narrow the window with a smaller "since".

Inputs: since, target, event_type.

system_infoB
Read-only

[READ] Daemon info: container/image counts, storage driver, kernel, resources.

Args: target: Target name from config; omit for the default.

Inputs: target.

system_versionB
Read-only

[READ] Docker version details (API version, Go version, components).

Args: target: Target name from config; omit for the default.

Inputs: target.

undo_applyA
Destructive

[WRITE][risk=medium] Apply a recorded undo by dispatching its inverse tool.

The inverse runs through its own governed tool, so it is audited on the same path as any other call. Pass dry_run=True to preview the inverse call without executing it. A token can only be applied once.

Args: undo_id: The undoId from undo_list (or an _undo_id in a write result). dry_run: If True, preview the inverse tool + params without running it. target: Passed through to the inverse tool when it accepts a target.

Inputs: undo_id (required), target, dry_run.

undo_listA
Read-only

[READ] List recorded, not-yet-applied undo tokens (most recent first).

Each entry names the original tool, the inverse tool that undo_apply would run, and a human note. Use the undoId with undo_apply.

Returns {"undos": [...], "returned": N, "limit": L, "truncated": bool}. truncated is measured (one extra row is fetched), not guessed from a length coincidence: when it is true there are MORE tokens than shown, so re-run with a higher limit rather than reporting the list as complete.

Each entry carries effectVerified. False means the original write lost its response, so the change it reverses is PROBABLE, not confirmed — check the live state before applying, and do not report the result as a restore of a state that may never have been reached.

Args: limit: Max rows to return (default 50, capped at 500). target: Unused (undo state is host-local); accepted for CLI uniformity.

Inputs: limit, target.

update_containerA
Destructive

[WRITE][risk=medium] Update a container's resource limits, capturing prior limits.

Captures the current CPU/memory HostConfig limits before the change so the harness records an undo (restore the prior limits). Pass dry_run=True to preview.

Args: container_id: Container id or name. resources: Resource limits to set — allowed keys: Memory, MemorySwap, MemoryReservation, NanoCpus, CpuQuota, CpuPeriod, CpuShares, CpusetCpus, CpusetMems (Docker update fields). dry_run: If True, preview without changing. target: Target name from config; omit for the default.

Inputs: resources (required), container_id (required), target, dry_run.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 38 tools

    Disambiguation4/5

    Most tools target a distinct resource+action (list/inspect/logs/stats per container, image, volume, network), so selection is generally clear. However there are real overlaps: restart_loop_rca vs container_restart_summary (both rank restart/exit-code), image_and_volume_bloat vs dangling_images/dangling_volumes/system_df, and overview vs system_info/system_version/system_df. Descriptions disambiguate most of these, but an agent could still hesitate between the aggregate and the granular tools.

    Naming Consistency3/5

    There is a recognizable verb_noun convention (list_containers, inspect_image, start_container, prune_volumes, update_container), but a large minority of tools are noun-first or noun-phrase (container_logs, container_stats, container_top, system_info, system_df, stack_detail, image_disk_usage, resource_pressure_analysis, restart_loop_rca, dangling_images). The mix is readable but not predictable enough for a clean 4.

    Tool Count3/5

    38 tools is on the heavy side (25+ is flagged as too many), and the surface spans containers, images, volumes, networks, pods, endpoints, stacks, system, undo, and several RCA analyzers. Each tool does map to something concrete, but the aggregate tools (overview, image_and_volume_bloat, restart_loop_rca) duplicate granular ones, so a few could be merged.

    Completeness4/5

    Coverage is strong for an AIOps/remediation server: full read+lifecycle over containers (list/inspect/logs/stats/top/start/stop/restart/remove/update), images, volumes, networks, pods, endpoints, stacks, system, plus undo and RCA analyses. Gaps exist (no container exec/create, no image pull/build, no volume/network create), but these may be intentionally out of scope for host triage.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      B
      maintenance
      Governed MySQL + MariaDB DBA operations — slow-query, lock-wait/deadlock, replication, and fragmentation RCA, with unbypassable audit logging (MCP + CLI), budget/runaway guards, dry-run, and undo/rollback.
      35
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Governed Prometheus + Grafana operations — firing-alert and scrape-target RCA, alert noise/flapping analysis, silences, and dashboards, with unbypassable audit logging (MCP + CLI), budget/runaway guards, dry-run, and undo/rollback.
      39
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Governed SSO/IAM operations for Keycloak and Authentik — login-failure, stale-permission, client-config, and MFA RCA, with unbypassable audit logging (MCP + CLI), budget/runaway guards, dry-run, and undo/rollback.
      29
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Governed CI/CD operations for self-hosted GitLab and Gitea — pipeline-failure, runner, artifact-bloat, and stale-branch RCA, with unbypassable audit logging (MCP + CLI), budget/runaway guards, dry-run, and undo/rollback.
      28
      MIT