Tuning Engines CLI, MCP server, and Python agent runtime adapters for governed model, agent, skill, and MCP workflows. Fine-tune open-source LLMs, run inference, manage datasets/evaluations, and connect LangGraph or Temporal while Tuning Engines handles policy, audit, usage, and token economics.
Tuning Engines is a governed AI runtime for model, agent, skill, and MCP workflows. Route inference through one OpenAI-compatible API, apply RBAC and traffic policies, request approvals for high-risk actions, inspect traces and usage, and connect durable orchestration frameworks such as LangGraph and Temporal. The same CLI and MCP server also manage domain-specific fine-tuning of open-source models.
Training Agents
Tuning Engines uses specialized agents that control how your data is analyzed and converted into training data. Each agent produces a different kind of domain-specific fine-tuned model optimized for its use case. Current agents focus on code, with more coming for customer support, data extraction, security review, ops, and other domains.
Cody (code_repo) — Code Autocomplete Agent
Cody fine-tunes on your GitHub repo using QLoRA (4-bit quantized LoRA) via the Axolotl framework (HuggingFace Transformers + PEFT). It learns your codebase's patterns, naming conventions, and project structure to produce a fast, lightweight adapter optimized for real-time completions.
te jobs create --agent code_repo \
--base-model Qwen/Qwen2.5-Coder-7B-Instruct \
--repo-url https://github.com/your-org/your-repo \
--output-name my-cody-model
SIERA (sera_code_repo) — Bug-Fix Specialist
SIERA (Synthetic Intelligent Error Resolution Agent) uses the Open Coding Agents approach from AllenAI to generate targeted bug-fix training data from your repository. It synthesizes realistic error scenarios and their resolutions, then fine-tunes a model that learns your team's debugging style, error handling conventions, and fix patterns.
tuningengines-cli · DSH Hub
Best for: debugging, error resolution, patch generation, root cause analysis, fix suggestions.
te jobs create --agent sera_code_repo \
--quality-tier high \
--base-model Qwen/Qwen2.5-Coder-7B-Instruct \
--repo-url https://github.com/your-org/your-repo \
--output-name my-siera-model
Quality tiers (SIERA only):
low — Faster, fewer synthetic pairs (default)
high — Deeper analysis, more training data, better results
Coming Soon
Agent
Persona
What it does
Resolve
Mira
Fine-tunes on support tickets, macros, and KB articles for automated ticket resolution
Extractor
Flux
Trains for strict schema extraction from docs, PDFs, and business text
Guard
Aegis
Security-focused code reviewer that catches risky patterns and proposes safer fixes
OpsPilot
Atlas
Incident response agent trained on runbooks, postmortems, and on-call notes
npm install -g tuningengines-cli
# Or run without installing
npx -y --package tuningengines-cli@latest te auth status
# Sign up or log in (opens browser — works for new accounts too)
te auth login
# Add credits (opens browser to billing page)
te billing add-credits
# Estimate cost before training
te jobs estimate --base-model Qwen/Qwen2.5-Coder-7B-Instruct
# Train Cody on your repo
te jobs create --agent code_repo \
--base-model Qwen/Qwen2.5-Coder-7B-Instruct \
--repo-url https://github.com/your-org/your-repo \
--output-name my-model
# Monitor training
te jobs status --watch
# View your trained models
te models list
# Create a governed orchestration starter
te orchestration init langgraph
te orchestration init temporal
te orchestration init inngest
te orchestration init triggerdev
te orchestration init hatchet
te orchestration init restate
te orchestration init dbos
te orchestration init dapr
te orchestration init prefect
te orchestration init dagster
te orchestration init airflow
MCP Server Setup
The CLI includes a built-in MCP server with 60+ tools. Any AI assistant that supports MCP can fine-tune models, manage training jobs, run evaluations, check inference usage, inspect traces, review approvals, and manage non-secret tenant registry metadata through natural language.
For security, the MCP server intentionally does not expose internal proxy routes. It also refuses MCP-side inference-key creation and raw secret-bearing mutation fields. Use the CLI or web UI for workflows that intentionally create one-time keys, submit raw provider secrets, validate S3 credentials, or import/export S3 assets with raw credentials.
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
claude mcp add tuning-engines -- npx -y --package tuningengines-cli@latest te mcp serve
Work Sessions and outcomes
Label the desired outcome for a project without interrupting your coding workflow:
te goal start "Fix flaky checkout retries"
te goal show
te goal complete --result succeeded
Install optional native telemetry hooks for Claude Code or Codex:
te guard claude-code install --mode observe --project .
te guard claude-code doctor
te guard claude-code doctor --probe
te guard codex install
te guard codex doctor
Claude Code writes project-local hooks into .claude/settings.local.json. On
Windows, verify with dir .\.claude, type .\.claude\settings.local.json,
then restart Claude Code from the same project root and review claude /hooks.
doctor --probe is available in tuningengines-cli 0.4.20 and later; it runs
synthetic hook events through the installed commands and checks that the trace is
visible to Tuning Engines. Hook invocations also write a local redacted status
log at .claude/tuning-engines-hook-status.jsonl.
Codex project hooks require review and trust from /hooks. The installer pins
hooks to the exact CLI executable that created them; te guard codex doctor
reports that path, CLI version, native-event contract, required lifecycle
hooks, and any conflicting te command found on PATH. Each Codex or Claude
Code user prompt starts a new trace while retaining one stable Work Session for
the native conversation. Tuning Engines sends pseudonymous session and
transcript references by default, not transcript contents or local absolute
paths.
Claude Code Plugin
The repository also ships a Claude Code plugin wrapper around the same MCP
server. It keeps installation discoverable while preserving the same
TE_API_KEY environment-variable boundary:
claude plugin marketplace add cerebrixos-org/tuning-engines-cli
claude plugin install tuning-engines@tuning-engines
The same plugin directory includes a Codex plugin manifest, marketplace entry,
MCP server, and tenant-operations skill. It uses the Tuning Engines web app as
the shared control-plane UI for traces, approvals, policies, cost, and review.
DeepSeek Harness Plugin
Install the native Harness bundle to capture turn, model-step, and tool
lifecycle telemetry and optionally enforce TE governance before tool execution:
"Fine-tune Qwen 7B on my-org/my-repo using the SIERA agent with high quality"
"How much would it cost to train a 32B model for 3 epochs on this repo?"
"Check the status of my latest training job"
"List my trained models"
"Export my model to s3://my-bucket/models/"
"Show my account balance"
"Train a bug-fix specialist on this repo" (auto-selects SIERA)
"Create an autocomplete model for this codebase" (auto-selects Cody)
The create_job tool description includes full agent details and model lists, so AI assistants automatically select the right agent and model based on what you ask for.
Unified API Endpoint
Tuning Engines can be used anywhere a tool accepts an OpenAI-compatible API
base URL. Point the client at:
https://api.tuningengines.com/v1
Use an inference key that starts with sk-te-... for live model calls, and use
the model IDs shown by:
te inference models
This lets OpenCode, Temporal activities, LangGraph apps, OpenAI SDK clients,
and other custom-provider clients route through the same Tuning Engines control
plane for model RBAC, routing, fallbacks, guardrails, AGT policy, traces,
usage metering, and cost attribution.
See docs/unified-api-endpoint.md for copy-paste
examples for OpenCode, Temporal, Python, JavaScript, and other
OpenAI-compatible clients.
Agent Runtime SDK and Orchestration Starters
Use the CLI/MCP package when you want npx tools for assistants. Use the
Python SDK when you want your own app to run durable agent workflows while
Tuning Engines remains the governed control plane for models, agents, skills,
MCP tools, RBAC, AGT policy, audit, usage, and token economics.
OpenAI Agents SDK users can install the native model and trace integration:
pip install "tuning-engines[openai-agents]"
See packages/tuning-agents/README.md
for setup. The OpenAI Agents runtime owns the loop; TE supplies the governed
endpoint and receives metadata-only SDK traces through its background processor.
The package installs the tuning_agents Python module.
LangGraph example:
from langgraph.checkpoint.memory import InMemorySaver
from tuning_agents import TuningClient
from tuning_agents.langgraph import create_tuning_langgraph_agent, invoke_with_trace
client = TuningClient(api_key="te_your_key_here")
agent = create_tuning_langgraph_agent(
client,
model="llama-3.3-70b-fp8",
agent_names=["billing-escalation"],
checkpointer=InMemorySaver(),
interrupt_before=["tools"],
)
result = invoke_with_trace(
client,
agent,
[{"role": "user", "content": "Triage this ticket and escalate if needed."}],
thread_id="ticket-123",
)
client.flush_trace(name="ticket-triage", runtime="langgraph", status="succeeded")
Temporal example:
from tuning_agents.temporal import (
TuningEnginesTemporalFeatures,
create_tuning_engines_plugin,
define_temporal_workflow,
)
plugin = create_tuning_engines_plugin(
features=TuningEnginesTemporalFeatures(
built_in_workflow=False,
model_calls=True,
skill_tools=True,
mcp_tools=True,
agents=True,
approvals=True,
traces=True,
state_references=True,
interventions=True,
model_catalog=True,
usage=True,
)
)
TuningAgentWorkflow = define_temporal_workflow()
# Pass plugin to Client.connect(..., plugins=[plugin]) and register the workflow.
The SDK captures runtime events from LangGraph/Temporal and posts them to
POST /api/v1/traces. Each event carries a run_id, request_id, and a
normalized event type such as model.call, mcp.tool_call, agent.message,
workflow.step, human.edit, action.finalized, outcome.recorded, or
state.reference. The app pairs that with inference usage, request capture,
policy decisions, approval requests, external state references, audit, and
billing logs.
JavaScript/TypeScript users can also import lightweight tracing helpers from
the npm package:
import { createOpenAIAgentsTraceAdapter } from "tuningengines-cli/adapters/openai-agents";
import { createClaudeAgentSdkTraceAdapter } from "tuningengines-cli/adapters/claude-agent-sdk";
Both helpers send redacted run, model, tool, handoff, error, goal, and outcome
events to the existing trace API. goal_key, goal_status, and goal_score
are normalized into the same success-signal analytics as outcome_key.
For decision traces, store redacted signals in metadata.decision, for example
proposal_summary, changed_fields, change_summary, final_action,
outcome_label, and reason_summary. Do not place raw prompts, provider keys,
tenant secrets, or full customer data in trace metadata.
Generate a starter kit:
te orchestration init langgraph --dir ./lg-te-demo
te orchestration init temporal --dir ./temporal-te-demo
te orchestration init inngest --dir ./inngest-te-demo
te orchestration init triggerdev --dir ./trigger-te-demo
te orchestration init hatchet --dir ./hatchet-te-demo
te orchestration init restate --dir ./restate-te-demo
te orchestration init dbos --dir ./dbos-te-demo
te orchestration init dapr --dir ./dapr-te-demo
te orchestration init prefect --dir ./prefect-te-demo
te orchestration init dagster --dir ./dagster-te-demo
te orchestration init airflow --dir ./airflow-te-demo
LangGraph and Temporal starters use the Python runtime SDK. Inngest,
Trigger.dev, and Hatchet starters generate TypeScript projects with a small
self-contained Tuning Engines helper. Restate, DBOS, and Dapr starters use the
same TypeScript helper. Prefect, Dagster, and Airflow starters generate Python
workflow examples with a small helper module. All generated examples include
governed model calls, trace flushing, registry manifests, policy context
metadata, decision metadata, runtime state references, and approval retry
patterns.
CLI Commands
Authentication
Command
Description
te auth login
Sign up or log in via browser
te auth logout
Clear saved credentials
te auth status
Show current auth status (email, balance)
Training Jobs
Command
Description
te jobs list
List all training jobs
te jobs show
Show job details
te jobs create
Submit a training job (--agent, --quality-tier, --base-model, --repo-url, --output-name)
te jobs status
Live status (--watch for continuous polling)
te jobs cancel
Cancel a running job
te jobs retry
Retry from last checkpoint
te jobs estimate
Cost estimate before submitting
te jobs validate-s3
Pre-validate S3 credentials
Models
Command
Description
te models list
List your trained models
te models show
Show model details
te models base
List supported base models
te models import
Import a model from S3
te models export
Export a model to S3
te models delete
Delete a model
te models status
Check import/export status
Datasets
Command
Description
te datasets list
List all datasets
te datasets show
Show dataset details
te datasets create
Create a dataset from S3 (--name, --s3-url, --for-evaluation)
te datasets delete
Delete a dataset
te datasets status
Check import/processing status
Evaluations
Command
Description
te evals list
List all evaluations
te evals show
Show evaluation details and scores
te evals create
Run an evaluation (--model, --dataset, --evaluators)
te evals cancel
Cancel a running evaluation
te evals status
Live evaluation progress
te evals evaluators
List available evaluators
te evals estimate
Cost estimate for an evaluation
Inference
Command
Description
te inference models
List available inference models
te inference usage
Show inference API usage stats
te inference jwt
Get a JWT for direct API access
te inference token
Exchange an inference key (sk-te-...) for a short-lived inference JWT
te inference chat --data ''
Run an OpenAI-compatible chat completion
te inference responses --data ''
Run an OpenAI Responses request
te inference embeddings --data ''
Create embeddings
te inference messages --data ''
Run an Anthropic-compatible Messages request
Runtime Traces and Approvals
Command
Description
te traces list
List LangGraph, Temporal, and custom runtime traces
te traces show
Show one trace, including events, policy decisions, and approvals when linked
te traces ingest --data ''
Ingest or update a trace using a user API token or inference key
te outcomes list
List observed outcomes, goals, evals, and workflow success signals
te outcomes record --run-id ... --key ... --label ...
Record a success signal for a run
te outcomes map --outcome-key ... --criteria ''
Map unmapped events to an outcome key
te insights list
List Insight Loop recommendations
te insights accept
Accept an insight as valid; does not change production
te insights apply
Apply or queue the approved action for an accepted insight
te doctor simulate --data ''
Simulate inference access, role, endpoint, policy, and resource checks
te policy-decisions list
List AGT YAML policy decisions
te policy-decisions show
Show one policy decision with redacted context
te policy-templates list
List curated AGT YAML policy templates
te policy-templates render --params ''
Render disabled/shadow policy YAML from safe structured parameters
te policy-drafts generate --prompt ''
Generate an AI-assisted disabled/shadow draft for review and testing
te approvals list --status pending
List policy approval requests
te approvals show
Show approval detail and retry metadata
te approvals approve
Approve a pending request
te approvals deny
Deny a pending request
te interventions list/show/request/ack/complete/fail
Manage runtime pause, resume, cancel, and replay requests
te state list/show/upsert
Manage safe external workflow-state and memory references
te registry sync --file tuning-registry.yml --dry-run/--apply
Diff or apply agent, skill, and MCP registry manifests
te registry show
Inspect an applied registry sync
te work-sessions list/show/complete/confirm-outcome
Inspect and update Work Sessions
te initiatives list/show/create/update
Manage strategic initiative groupings
Orchestration Starters
Command
Description
te orchestration init langgraph
Create a LangGraph starter wired to Tuning Engines governance and traces
te orchestration init temporal
Create a Temporal worker starter wired to Tuning Engines governance and traces
te orchestration init inngest
Create an Inngest function starter wired to Tuning Engines governance and traces
te orchestration init triggerdev
Create a Trigger.dev task starter wired to Tuning Engines governance and traces
te orchestration init hatchet
Create a Hatchet workflow starter wired to Tuning Engines governance and traces
te orchestration init restate
Create a Restate service starter wired to Tuning Engines governance and traces
te orchestration init dbos
Create a DBOS workflow starter wired to Tuning Engines governance and traces
te orchestration init dapr
Create a Dapr Workflow starter wired to Tuning Engines governance and traces
te orchestration init prefect
Create a Prefect flow starter wired to Tuning Engines governance and traces
te orchestration init dagster
Create a Dagster asset starter wired to Tuning Engines governance and traces
te orchestration init airflow
Create an Airflow DAG starter wired to Tuning Engines governance and traces
Agents
Command
Description
te agents list
List available agents
te agents show
Show agent details and capabilities
te agents message --data ''
Send a governed A2A agent message
Skills and MCP Execution
Command
Description
te skills list
List skills visible to the inference identity
te skills prepare --data ''
Prepare a governed skill
te skills invoke --data ''
Invoke a governed skill
te mcp call --server --tool --arguments ''
Call an enabled governed MCP tool
te mcp rediscover
Refresh MCP tool discovery
te mcp tools list/enable/disable/toggle
Administer discovered MCP tools
Compliance Automation
Command
Description
te compliance validate --data ''
Validate bounded content against adopted rulepacks
te compliance rewrite --data ''
Produce and revalidate a safe rewrite
te compliance evidence
Inspect compliance evidence
te compliance risks list/show/create/update
Manage the tenant risk register
te compliance risks assess/map-control/add-subject/remove-subject
Operate the risk-to-control lifecycle
te compliance source-runs create/show/submit-results/complete
Ingest normalized scanner or webhook results
te compliance certifications create/show
Run and inspect compliance certification jobs
Tenant Admin Automation
These commands require an API token for a tenant owner or tenant admin. They are
designed for CI smoke tests and end-to-end product checks. Secret fields can be
sent on create/update where the server supports them, but responses never print
stored provider keys, AWS secrets, or invitation tokens.
Command
Description
te tenant resources
List supported tenant resource names
te tenant list
List resources such as inference_keys, inference_roles, model_deployments, routing_profiles, guardrail_policies, governance_policies, mcp_servers, tenant_agents, tenant_skills, and credential_sources
te tenant show
Show one tenant resource
te tenant create --data ''
Create a tenant resource from JSON
te tenant update --data ''
Update a tenant resource from JSON
te tenant delete
Delete a tenant resource; inference keys are revoked
te tenant validate guardrail_policies --data '' --sample-text 'hello'
Validate/test an unsaved simple guardrail without creating records
te tenant validate governance_policies --data '' --context ''
Validate/test an unsaved Governance Rule without creating records
te tenant test-policy --context ''
Dry-run a Governance Rule
te tenant test governance_policies --context ''
Compatibility alias for governance policy dry-runs
te tenant team list
List tenant members, pending invitations, and allowed domains
te tenant team invite --role member
Invite a user by email; the invite token is emailed and never printed
te tenant team set-role --inference-role-id
Assign an inference role to a member
te tenant team disable
Disable a member
te tenant team enable
Re-enable a member
te tenant team remove
Remove a member
te tenant team cancel-invite
Cancel a pending invitation
te tenant team domains --set "example.com,example.org"
Replace allowed email domains
te tenant capture show
Show inference capture settings
te tenant capture update --data ''
Update inference capture settings
Billing & Account
Command
Description
te billing show
Balance and transaction history
te billing add-credits
Open browser to add credits
te account
Account info
Configuration
Command
Description
te config set-token
Set API key manually
te config set-url
Override API URL
te config show
Show current config
All commands support --json for machine-readable output.
MCP Tools Reference
Training Jobs
Tool
Description
create_job
Fine-tune an LLM on a GitHub repo. Supports agent selection (Cody, SIERA), quality tier, base model, epochs, S3 export.
estimate_job
Cost estimate before training. Returns cost range, balance, sufficiency check.
list_jobs
List training jobs with status filter
show_job
Full job details including agent, model, GPU usage, cost, retry info
job_status
Live status with GPU minutes, charges, delivery progress
cancel_job
Cancel a running/queued job
retry_job
Retry a failed job from its last checkpoint
Models
Tool
Description
list_models
List trained and imported models
show_model
Model details (status, size, base model, training job)
delete_model
Delete a model from cloud storage
model_status
Import/export progress
list_supported_models
Available base models with GPU hours per epoch
Marketplace
Tool
Description
list_catalog_models
Browse pre-built models and datasets
get_catalog_model
Details of a marketplace item
catalog_export_status
Check marketplace export progress
Datasets
Tool
Description
list_datasets
List datasets for training and evaluation
show_dataset
Dataset details and status
create_dataset
Create a dataset from S3
delete_dataset
Delete a dataset
dataset_status
Check dataset import/processing status
Evaluations
Tool
Description
list_evaluations
List model evaluations
show_evaluation
Evaluation details, scores, and metrics
create_evaluation
Run an evaluation against a dataset
cancel_evaluation
Cancel a running evaluation
evaluation_status
Live evaluation progress
list_evaluators
Available evaluators (code_execution, similarity, llm_judge, etc.)
estimate_evaluation
Cost estimate for an evaluation
Inference
Tool
Description
list_inference_models
Models available for inference
inference_usage
Inference API usage statistics
get_inference_jwt
Get JWT token for direct API access
get_inference_token
Exchange an inference key for a short-lived inference JWT
call_inference
Call chat, Responses, embeddings, or Messages using the configured credential
send_agent_message
Send a governed A2A agent message
list_skills / invoke_skill
Discover, prepare, or invoke governed skills
Runtime, Policy, and Approvals
Tool
Description
list_traces
List runtime traces
show_trace
Show a trace with linked events, policy decisions, and approvals
create_trace
Ingest a trace payload without secrets
list_outcomes
List observed outcomes/goals normalized as success signals
list_insights
List Insight Loop recommendations
show_insight
Show one Insight Loop recommendation
doctor_simulate
Simulate inference access, role, endpoint, policy, and resource checks
record_outcome
Record an outcome/goal signal; requires --enable-registry-writes
map_outcome
Create an outcome mapping rule; requires --enable-registry-writes
accept_insight
Accept an insight for review; requires --enable-registry-writes
apply_insight
Apply or queue an accepted insight; requires --enable-registry-writes
list_policy_decisions
List AGT YAML policy decisions
show_policy_decision
Show one decision with redacted context
list_policy_templates
List curated AGT YAML policy templates
render_policy_template
Render disabled/shadow policy YAML from safe structured parameters
generate_policy_draft
Generate an AI-assisted disabled/shadow draft; secret-looking prompts are refused
Ingest normalized external test results; writes require --enable-registry-writes
show_compliance_certification
Inspect a certification run
Tenant Admin MCP Tools
These tools require a tenant owner/admin API token. The MCP server refuses internal
proxy routes, inference-key creation, and raw secret-bearing mutation fields.
Tool
Description
list_tenant_resources
List allowlisted tenant resource names
tenant_resource_list
List models, roles, policies, MCP servers, agents, skills, credential sources, and related metadata
tenant_resource_show
Show one resource without returning stored secrets
tenant_resource_create
Create non-secret tenant registry/config metadata
tenant_resource_update
Update non-secret tenant registry/config metadata
tenant_resource_delete
Delete or revoke a tenant resource
tenant_resource_validate
Validate/test unsaved guardrail or AGT policy payloads without creating records
test_governance_policy
Dry-run an AGT YAML governance policy
tenant_team_list
List members, invitations, and allowed domains
tenant_team_invite
Invite a user without returning invitation tokens
tenant_team_set_inference_role
Assign or clear an inference role
tenant_team_disable / tenant_team_enable
Disable or re-enable a member
tenant_team_remove
Remove a tenant member
tenant_invitation_cancel
Cancel a pending invitation
tenant_domains_update
Replace allowed email domains
inference_capture_show / inference_capture_update
Manage request-capture settings using credential-source references
Agents
Tool
Description
list_agents
List available agents
show_agent
Agent details and capabilities
Account
Tool
Description
get_balance
Account balance and recent transactions
get_account
Account details
Environment Variables
Variable
Description
TE_API_KEY
API key (overrides config file)
TE_API_URL
API URL (default: https://app.tuningengines.com)
TE_INFERENCE_KEY
Optional sk-te-* key for direct model, agent, skill, and MCP calls
TE_INFERENCE_URL
Inference base URL (default: https://api.tuningengines.com/v1)
Tenant management commands keep the configured te_* API token local and
exchange it for a short-lived management JWT before calling the API. Inference
keys (sk-te-*) are for inference-only flows such as te inference token and
proxy calls; they are not accepted for tenant registry management commands.
Inference Smoke Testing
Use te-inference-smoke to exercise inference behavior as a tenant admin and, optionally, real tenant users. The default run is read-only. Set TE_SMOKE_MUTATE=1 to create temporary inference roles, keys, policies, guardrails, MCP servers, agents, and skills, then test permission permutations and clean them up.
If you only have an sk-te-* inference key, set TE_INFERENCE_KEY for
proxy-only checks. Full role/user/policy permutations require a tenant-admin
app API key that starts with te_.
TE_SMOKE_CREATE_MODEL_DEPLOYMENT=1 is useful for disposable tenants that do
not already have an enabled model. By default the runner treats a provider
authentication failure on an allowed model as proof that Tuning Engines RBAC
allowed the request through to the provider. Set
TE_SMOKE_ALLOW_PROVIDER_AUTH_FAILURE=0 when the tenant has real provider
credentials and the allowed call must return 200.
To test multiple tenant users, provide their API tokens:
