> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usesynth.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Tool Reference

> Managed Research MCP tools: full smr_* inventory grouped by intent.

Managed Research MCP tools mostly use stable **`smr_*`** names. Synth Tag beta
tools intentionally use **`tag_*`** names so delegate sessions do not collide
with lower-level Factory and run-control tools. Use
[MCP Quickstart](/managed-research/mcp-quickstart) for setup and
[Get started](/managed-research/get-started) for Codex and Claude Code install.

The authoritative enumeration is **`ManagedResearchMcpServer._build_tools`** in [`synth_ai/managed_research/mcp/server.py`](https://github.com/synth-laboratories/synth-ai/blob/dev/synth_ai/managed_research/mcp/server.py) in the **`synth-ai`** package. The lists below track that surface; if a client is missing a tool, upgrade the package.

## Synth Tag

* `tag_create_session`
* `tag_get_session`
* `tag_send_message`

See [Synth Tag](/managed-research/tag) for the SDK and MCP delegate flow.

## Health, capabilities, and billing

* `smr_health_check`
* `smr_get_capabilities`
* `smr_get_limits`
* `smr_get_capacity_lane_preview`
* `smr_get_billing_entitlements`

## Projects

* `smr_create_runnable_project`
* `smr_list_projects`
* `smr_get_project`
* `smr_get_default_project`
* `smr_rename_project`
* `smr_patch_project`
* `smr_get_project_status`
* `smr_get_project_workspace`
* `smr_pause_project`
* `smr_resume_project`
* `smr_archive_project`
* `smr_unarchive_project`
* `smr_get_project_entitlement`
* **Notes and knowledge:** `smr_get_project_notes`, `smr_set_project_notes`, `smr_append_project_notes`, `smr_get_org_knowledge`, `smr_set_org_knowledge`, `smr_get_project_knowledge`, `smr_set_project_knowledge`, `smr_curated_knowledge`

## Changesets

* `smr_list_project_changesets`
* `smr_create_project_changeset`
* `smr_get_project_changeset`
* `smr_decide_project_changeset`

## Provider keys

* `smr_set_provider_key`
* `smr_get_provider_key_status`

## GitHub setup

* `smr_setup_github_status`
* `smr_setup_github_start_oauth`
* `smr_setup_github_list_repos`
* `smr_setup_github_disconnect`

## Exports

* `smr_setup_exports_list_targets`
* `smr_setup_exports_create_target`

## Workspace and inputs

* `smr_attach_source_repo`
* `smr_get_workspace_inputs`
* `smr_upload_workspace_files`
* `smr_get_workspace_download_url`
* `smr_get_project_git`
* `smr_download_workspace_archive`

## Objectives and work-mode helpers

* `smr_open_ended_questions`
* `smr_objectives`
* `smr_directed_effort_outcomes`

## Progress, setup, launch, readiness

* `smr_get_project_setup`
* `smr_prepare_project_setup`
* `smr_list_runbook_presets`
* `smr_get_launch_preflight`
* `smr_status_readiness`

## Repos, datasets, and work files

* **Work repos:** `smr_work_repos_list`, `smr_work_repos_attach`, `smr_work_repos_detach`
* **Work datasets:** `smr_work_datasets_list`, `smr_work_datasets_upload`, `smr_work_datasets_download`
* **Work files:** `smr_work_files_list`, `smr_work_files_upload`
* **Run output files:** `smr_list_run_output_files`, `smr_get_run_output_file_content`

## Resources (catalog-style)

* **Project files:** `smr_list_project_files`, `smr_create_project_files`, `smr_get_project_file`, `smr_get_file_content`
* **External repositories:** `smr_list_project_external_repositories`, `smr_create_project_external_repository`, `smr_patch_project_external_repository`
* **Credential refs:** `smr_list_project_credential_refs`, `smr_create_project_credential_ref`, `smr_patch_project_credential_ref`
* **Run mounts and bindings:** `smr_list_run_file_mounts`, `smr_upload_run_files`, `smr_list_run_repository_mounts`, `smr_create_run_repository_mount`, `smr_list_run_credential_bindings`, `smr_create_run_credential_binding`

## Runs and control

* **Start and list:** `smr_trigger_run`, `smr_start_one_off_run`, `smr_list_runs`, `smr_get_run`, `smr_list_active_runs`
* **Lifecycle:** `smr_stop_run`, `smr_pause_run`, `smr_resume_run`, `smr_branch_run_from_checkpoint`
* **Contract and execution:** `smr_get_run_contract`, `smr_get_run_execution`
* **Events and graph:** `smr_list_run_task_events`, `smr_list_run_objective_events`, `smr_get_run_work_graph`, `smr_get_run_logical_timeline`, `smr_get_run_event_log`
* **Authority and evidence:** `smr_get_run_authority_readouts`, `smr_get_run_operator_evidence`
* **Traces:** `smr_get_run_traces`, `smr_get_run_actor_trace`, `smr_list_run_actor_traces`, `smr_get_raw_trace_events`, `smr_download_raw_trace`
* **Actors and participants:** `smr_get_run_actor_usage`, `smr_control_project_run_actor`, `smr_list_run_participants`, `smr_get_run_artifact_progress`, `smr_list_run_actor_logs`, `smr_get_run_primary_parent`, `smr_run_objective_scopes`
* **Steering:** `smr_runtime_message_queue`, `smr_runtime_intents`
* **Transcript and questions:** `smr_get_run_transcript`, `smr_list_run_questions`
* **Checkpoints:** `smr_create_run_checkpoint`, `smr_list_run_checkpoints`, `smr_restore_run_checkpoint`

## Artifacts and run-scoped lists

* `smr_list_run_artifacts`
* `smr_get_run_artifact_manifest`
* `smr_get_artifact`
* `smr_get_artifact_content`
* `smr_download_artifact`
* `smr_list_run_models`
* `smr_list_run_datasets`

## Outputs and work products

* `smr_list_run_work_products`
* `smr_get_run_work_product`
* `smr_export_run_work_product`
* `smr_explain_work_product_blocker`
* **Container eval:** `smr_upload_container_eval_package`, `smr_validate_container_eval_package`

## Results: models and pull requests

* **Models:** `smr_results_models_list`, `smr_results_models_get`, `smr_results_models_download`, `smr_results_models_export`
* **PRs:** `smr_results_prs_list`, `smr_results_prs_get`

## Trained models (extended)

* `smr_register_trained_model`
* `smr_get_trained_model`
* `smr_list_trained_models_for_run`
* `smr_export_trained_model`
* `smr_create_trained_model_adapter_upload_url`
* `smr_complete_trained_model_adapter_upload`
* `smr_update_trained_model`
* `smr_delete_trained_model`
* `smr_get_run_cost_summary`
* `smr_report_tinker_training_usage`

## Approvals and questions

* `smr_list_run_approvals`
* `smr_approve_run_approval`
* `smr_deny_run_approval`
* `smr_respond_to_run_question`

## Usage and economics

* `smr_get_run_usage`
* `smr_get_run_resource_limits`
* `smr_get_run_progress_toward_resource_limits`
* `smr_get_project_usage`
* `smr_get_project_resource_limits`
* `smr_get_project_progress_toward_resource_limits`
* `smr_request_resource_limit_extension`

Limit progress tools return configured caps, usage, remaining budget, active
blockers, and extension posture. Orchestrators should use them to ration work
before a hard block: reduce fanout, choose a cheaper model, summarize context,
or request an extension early.

## Logs

* `smr_list_run_log_archives`

## Launch contract

`smr_start_one_off_run` and `smr_trigger_run` require a coherent launch payload (for example `work_mode` and `providers`). A typical public example:

```json theme={null}
{
  "host_kind": "daytona",
  "work_mode": "directed_effort",
  "providers": [{"provider": "openrouter"}],
  "runbook": "lite"
}
```

Launch denials surface as **MCP tool errors**. Treat them as failed tool calls and inspect structured error data when your client exposes it.