Tuning Engines - Governed AI Runtime MCP server

Read from the package source without running it. The installed server may list more.

• accept_insight

  Accept an Insight Loop recommendation as valid for review. Requires --enable-registry-writes. Does not change production.

• activate_context_asset

  Explicitly activate one reviewed context version. Requires --enable-registry-writes.

• apply_insight

  Apply or queue the approved action for an accepted Insight Loop recommendation. Requires --enable-registry-writes.

• apply_work_session_repair

  Apply a previously previewed Work Session repair.

• approve_approval

  Approve a pending policy approval request. Requires tenant owner/admin API token.

• cancel_jobwrite action

  Cancel a running or queued fine-tuning job. The job will be charged for any GPU time already used.

• complete_compliance_source_runwrite action

  Finalize an external compliance source run.

• complete_work_session

  Mark a Work Session completed. Requires --enable-registry-writes.

• create_compliance_source_runwrite action

  Create an external compliance source run. Requires owner/admin access.

• create_context_asset_draftwrite action

  Create a disabled context draft. It never activates automatically. Requires --enable-registry-writes.

• create_jobwrite action

  Fine-tune an LLM on a GitHub repository using Tuning Engines.

• create_runtime_interventionwrite action

  Request a pause, resume, cancel, or replay. Requires owner/admin access and --enable-registry-writes.

• create_tracewrite action

  Ingest or update a runtime trace. Include run_id/request_id and normalized event types when possible. metadata.decision must contain redacted summaries only. Do not include secrets. Works with a user API token or inference key.

• delete_modelwrite action

  Delete a trained model from cloud storage.

• deny_approval

  Deny a pending policy approval request. Requires tenant owner/admin API token.

• doctor_simulatewrite action

  Run an Inference Doctor simulation for access, role, endpoint, policy, and resource debugging.

• estimate_job

  Get a cost estimate for a fine-tuning job before submitting it. Returns estimated cost, cost range, current balance, and whether balance is sufficient. Always estimate before creating a job.

• flush_inference_capture

  Queue pending and failed request captures for delivery. Requires --enable-registry-writes.

• freeze_trajectory_evidence_set

  Freeze reviewed Work Sessions into an immutable evidence version. Requires --enable-registry-writes.

• generate_policy_draft

  Generate an AI-assisted AGT YAML draft. Drafts are disabled/shadow and must be reviewed/tested/saved explicitly. Do not include secrets.

• get_account

  Get your Tuning Engines account details and settings.

• get_balance

  Check your Tuning Engines account balance and recent transactions.

• inference_capture_show

  Show request-capture settings for fine-tuning data capture. Secret values are not returned.

• inference_capture_updatewrite action

  Update request-capture settings. Use credential_source_id references; raw cloud secrets are refused by MCP.

• install_mcp_template

  Install a verified MCP template as a disabled server. Uses an existing secret_reference_id.

• job_status

  Get live status of a fine-tuning job including current status, GPU minutes used, estimated charges, remaining balance, and delivery progress. Use this to monitor a running job.

• list_ai_system_assets

  List authorized AI system assets across agents, MCP, tools, skills, models, workflows, context, identities, and infrastructure.

• list_approvals

  List policy approval requests for the current tenant. Requires tenant owner/admin API token.

• list_compliance_risks

  List tenant compliance risks. Requires compliance reader access.

• list_context_assets

  List governed context assets. Content remains versioned and inactive until explicitly activated.

• list_inference_feedback

  List redacted inference feedback metadata for the current scope.

• list_initiatives

  List strategic initiatives that group Work Sessions.

• list_insights

  List Insight Loop recommendations.

• list_intelligence_runs

  List reproducible trajectory intelligence executions and their version/status metadata.

• list_jobs

  List fine-tuning training jobs on Tuning Engines. Returns recent jobs with status, base model, agent type, GPU usage, and cost. Use this to check on existing training runs or find a job ID.

• list_mcp_templates

  List verified MCP marketplace templates.

• list_models

  List your trained and imported models on Tuning Engines.

• list_outcomes

  List observed outcomes, goals, workflow statuses, evals, and feedback normalized into success signals.

• list_policy_decisions

  List AGT YAML policy decisions for the current tenant. Requires tenant owner/admin API token.

• list_policy_templates

  List curated AGT YAML policy templates. Requires tenant owner/admin API token.

• list_runtime_interventions

  List tenant-scoped pause, resume, cancel, and replay requests.

• list_runtime_state_references

  List safe pointers to external workflow state and memory. Memory content and secrets are never returned.

• list_supported_models

  List the supported base HuggingFace models available for fine-tuning on Tuning Engines. Optionally filter by agent to see only compatible models.

• list_tenant_resources

  List tenant-admin resource names available through the public API. Internal proxy routes are intentionally not exposed.

• list_traces

  List runtime traces emitted by LangGraph, Temporal, MCP, skills, agents, or custom runtimes. Requires a tenant user API token.

• list_trajectory_evidence_sets

  List immutable, reviewed evidence-set versions used by trajectory intelligence.

• list_work_sessions

  List tenant-scoped Work Sessions.

• map_outcomewrite action

  Create an outcome mapping rule for events that omitted outcome_key/goal_key. Requires --enable-registry-writes.

• preview_trajectory_evidence_set

  Preview selected Work Sessions and automatically correlated evidence without freezing or mutating records.

• preview_work_session_repair

  Preview the exact and automatically correlated evidence affected by a repair.

• propose_outcome

  Propose a governed outcome for admin review.

• record_inference_feedback

  Record bounded feedback linked by request_id/run_id. Secret-bearing payloads are refused.

• record_outcome

  Record an outcome or goal success signal for a run. Requires --enable-registry-writes because it writes telemetry.

• registry_sync_apply

  Apply a secret-free registry manifest. Requires --enable-registry-writes.

• registry_sync_dry_runwrite action

  Validate and diff a secret-free registry manifest without mutation.

• render_policy_template

  Render a curated AGT YAML policy template into disabled/shadow YAML. Template params must not include secrets.

• resolve_governed_context

  Resolve authorized, versioned Tuning Engines context for a goal or action. Observe mode returns only match lineage and never changes execution.

• retry_job

  Retry a failed fine-tuning job from its last checkpoint. Creates a new job that resumes training where the failed one stopped, saving GPU time. Each retry is billed separately.

• retry_tenant_resource_sync

  Retry synchronization for a model, MCP server, agent, or skill.

• review_context_asset

  Review a context draft. Deterministic candidates require evidence-backed release gates. Requires --enable-registry-writes.

• show_ai_system_asset

  Show one authorized asset and its reviewed current relationships.

• show_approval

  Show one policy approval request with redacted context. Requires tenant owner/admin API token.

• show_compliance_certification

  Show one compliance certification run.

• show_compliance_evidence

  Show one tenant-scoped compliance evidence decision.

• show_compliance_risk

  Show a compliance risk with resources, controls, findings, treatments, and assessments.

• show_context_asset

  Show one governed context asset and its immutable versions.

• show_initiative

  Show one strategic initiative.

• show_insight

  Show one Insight Loop recommendation.

• show_intelligence_runwrite action

  Show one intelligence run, including evidence digest, versions, safe result summary, cost, and latency.

• show_job

  Get full details of a specific fine-tuning job including status, base model, agent type, GPU minutes, cost, error messages, and whether it can be retried from checkpoint.

• show_mcp_template

  Show setup and risk metadata for one MCP template.

• show_model

  Get details of a specific trained model.

• show_policy_decision

  Show one policy decision with redacted context and metadata. Requires tenant owner/admin API token.

• show_registry_sync

  Show one registry-manifest sync result.

• show_runtime_intervention

  Show one tenant-scoped runtime intervention.

• show_runtime_state_reference

  Show one safe external state or memory reference.

• show_trace

  Show one runtime trace by run_id, including events, policy decisions, and approvals when linked.

• show_trajectory_evidence_set

  Show one immutable evidence set and its safe evidence locators.

• show_work_session

  Show one Work Session and its linked traces.

• start_intelligence_runwrite action

  Queue versioned trajectory intelligence. Suggestions remain review-only. Requires --enable-registry-writes.

• submit_compliance_source_resultswrite action

  Submit normalized external compliance results. Raw cloud credentials are refused.

• tenant_domains_updatewrite action

  Replace the tenant's allowed email domains. Pass an empty array to allow any domain.

• tenant_invitation_cancelwrite action

  Cancel a pending tenant invitation. Requires tenant owner/admin API token.

• tenant_resource_createwrite action

  Create an allowlisted tenant resource using non-secret JSON. MCP refuses raw secret fields and inference-key creation; use CLI/web UI for those.

• tenant_resource_deletewrite action

  Delete or revoke an allowlisted tenant resource. Requires tenant owner/admin API token.

• tenant_resource_list

  List an allowlisted tenant resource. Requires tenant owner/admin API token.

• tenant_resource_show

  Show one allowlisted tenant resource. Secret values are not returned by the API.

• tenant_resource_updatewrite action

  Update an allowlisted tenant resource using non-secret JSON. MCP refuses raw secret fields; use CLI/web UI for secret updates.

• tenant_resource_validate

  Validate an unsaved model, MCP server, agent, skill, guardrail, or governance policy without creating records.

• tenant_team_disable

  Disable a tenant member and block API access. Requires tenant owner/admin API token.

• tenant_team_enable

  Re-enable a disabled tenant member. Requires tenant owner/admin API token.

• tenant_team_invite

  Invite a tenant member by email. Invitation token is emailed by the app and is never returned.

• tenant_team_list

  List tenant members, pending invitations, and allowed email domains. Requires tenant owner/admin API token.

• tenant_team_removewrite action

  Remove a tenant member. Requires tenant owner/admin API token.

• tenant_team_set_inference_role

  Assign or clear an inference role for a tenant member. Requires tenant owner/admin API token.

• test_governance_policy

  Dry-run an AGT YAML governance policy against a JSON context. Requires tenant owner/admin API token.

• undo_work_session_repair

  Undo a Work Session repair.

• upsert_runtime_state_reference

  Upsert a safe external state/memory pointer. Raw memory content and secrets are refused.

• validate_compliance

  Evaluate bounded content against adopted compliance rulepacks without mutating policy configuration.

• verify_tenant_secret_reference

  Verify a vault connection or secret reference server-side without returning the secret.

• Code scan108 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitystatic API keys via environment variables6/15
• Maintenancelast push 37 days ago12/15
• Maintainer identityregistry namespace matches repository owner7/10

Safety score

77/100

higher than 50% of 4,460 AI and agents servers

Tools

100

the typical AI and agents server has 6

Tools that change things

21%

typical: 0%. Fewer write tools means less an agent can break

GitHub stars

6

more than 80% of 2,711 AI and agents servers

Downloads a week

81

more than 59% of 1,538 AI and agents servers with a package

Tuning Engines - Governed AI Runtime: common questions

Is Tuning Engines - Governed AI Runtime MCP server safe?

Mostly: it is graded B (77/100). Read the Tuning Engines - Governed AI Runtime safety report

How do I install Tuning Engines - Governed AI Runtime?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does Tuning Engines - Governed AI Runtime need an API key?

Yes. The registry entry asks for TE_API_KEY.

Is Tuning Engines - Governed AI Runtime maintained?

The last commit was 78 days ago (2026-08-18). The latest release is v0.4.34.

What can I use instead of Tuning Engines - Governed AI Runtime?

Servers from other publishers that do the same job: HOL Guard MCP server, Pm Claude Skills MCP server and Agent-Native Analytics MCP server.

How does Tuning Engines - Governed AI Runtime compare with other AI and agents MCP servers?

safety score 77/100 (higher than 50% of 4,460 AI and agents servers); tools 100 (the typical AI and agents server has 6); tools that change things 21% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.