# Synth AI

## Docs

- [Benchmark Improvement](https://docs.usesynth.ai/managed-research/benchmark-improvement.md): Run a benchmark-focused Managed Research task with evidence and budget controls.
- [Beta Access](https://docs.usesynth.ai/managed-research/beta-access.md): What Beta Access unlocks for Managed Research.
- [Checkpoint Branching](https://docs.usesynth.ai/managed-research/checkpoint-branching.md): Create checkpoints and branch a run when a new direction is needed.
- [Concepts](https://docs.usesynth.ai/managed-research/concepts.md): Projects, runs, workers, context, artifacts, budgets, and approvals.
- [Data Factory Flow](https://docs.usesynth.ai/managed-research/data-factory-flow.md): Use Managed Research data-factory tools for intake, generation, review, and publish flows.
- [Eval Harness Improvement](https://docs.usesynth.ai/managed-research/eval-harness-improvement.md): Use Managed Research to improve an eval harness and leave reviewable evidence.
- [Get started](https://docs.usesynth.ai/managed-research/get-started.md): Install synth-ai[research], connect the Managed Research MCP, and verify connectivity.
- [Managed Research](https://docs.usesynth.ai/managed-research/intro.md): Run hosted AI workers against repos and research goals, then inspect durable evidence.
- [Launch Fields](https://docs.usesynth.ai/managed-research/launch-fields.md): Reference for the main fields that shape a Managed Research run.
- [MCP Quickstart](https://docs.usesynth.ai/managed-research/mcp-quickstart.md): Connect a hosted MCP client, start a Managed Research run, and inspect evidence.
- [MCP Tool Reference](https://docs.usesynth.ai/managed-research/mcp-reference.md): Managed Research MCP tools: full smr_* inventory grouped by intent.
- [Models and Harnesses](https://docs.usesynth.ai/managed-research/models-and-harnesses.md): Public Managed Research harnesses, model IDs, and reasoning-effort support.
- [OpenCode Harness](https://docs.usesynth.ai/managed-research/opencode-harness.md): Pin the OpenCode SDK harness and a supported model.
- [Preflight and Errors](https://docs.usesynth.ai/managed-research/preflight-and-errors.md): Diagnose launch blockers, preflight denials, budget stops, and unsupported configuration.
- [Projects and Context](https://docs.usesynth.ai/managed-research/projects-and-context.md): Use projects to bind repos, files, datasets, credentials, notes, and reusable knowledge.
- [Providers and Credentials](https://docs.usesynth.ai/managed-research/providers-and-credentials.md): Bind provider capability for Managed Research launches without putting secrets in prompts.
- [Managed Research Quickstart](https://docs.usesynth.ai/managed-research/quickstart.md): Start one hosted research run, then read back state and artifacts.
- [Repo Review and PR](https://docs.usesynth.ai/managed-research/repo-review-and-pr.md): Ask Managed Research to review a repo, make a focused change, and return PR-ready evidence.
- [Research Factory](https://docs.usesynth.ai/managed-research/research-factory.md): Run longer-lived research programs on the same SynthClient().research and MCP surfaces as Managed Research.
- [Runbooks and Work Modes](https://docs.usesynth.ai/managed-research/runbooks-and-work-modes.md): Choose run collaboration posture and goal posture for Managed Research.
- [Runs and Evidence](https://docs.usesynth.ai/managed-research/runs-and-evidence.md): Inspect run state, messages, tasks, actors, artifacts, checkpoints, usage, reports, and PRs.
- [Managed Research Scenarios](https://docs.usesynth.ai/managed-research/scenarios.md): Canonical scenarios used to validate Managed Research end-to-end.
- [Python SDK Reference](https://docs.usesynth.ai/managed-research/sdk.md): SynthClient().research namespaces, install, and launch patterns.
- [Python SDK Quickstart](https://docs.usesynth.ai/managed-research/sdk-quickstart.md): Use SynthClient().research to start runs, preflight projects, and inspect evidence.
- [Synth Tag](https://docs.usesynth.ai/managed-research/tag.md): Delegate a one-off research task, steer the active run, and retrieve a terminal run receipt.
- [Upgrading to synth-ai 0.11.2](https://docs.usesynth.ai/managed-research/upgrading.md): What changed in the Managed Research SDK between 0.11.1 and 0.11.2, and what to use instead.
- [Usage, Plans, and Budgets](https://docs.usesynth.ai/managed-research/usage.md): Track Managed Research plan state, included usage, flex credits, reset grants, promos, and budget caps.
- [Optimize Your Model's Prompts using GEPA](https://docs.usesynth.ai/prompt-optimization-gepa.md): Full walkthrough of GEPA prompt optimization with Synth AI
- [Overview](https://docs.usesynth.ai/prompt-optimization/overview.md): Optimize prompts using GEPA with verifier-guided evaluation. Works with any language via OpenAPI contracts.
- [ContainersClient](https://docs.usesynth.ai/reference/sdk/containers-client.md)
- [SDK Reference](https://docs.usesynth.ai/reference/sdk/index.md): Auto-generated Python SDK reference for SynthClient and SynthClient().research.
- [ContainerPoolsClient](https://docs.usesynth.ai/reference/sdk/pools-client.md)
- [AsyncResearchClient](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-async_client.md)
- [ResearchClient](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-client.md)
- [Enums](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-enums.md)
- [Errors](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-errors.md)
- [Factories](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-factories.md)
- [Limits](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-limits.md)
- [Models](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-models.md)
- [Project namespaces](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-project_namespaces.md)
- [Projects](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-projects.md)
- [Run readouts](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-run_readouts.md)
- [Runs](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-runs.md)
- [Secrets](https://docs.usesynth.ai/reference/sdk/research/synth_ai-research-secrets.md)
- [SynthClient](https://docs.usesynth.ai/reference/sdk/synth-client.md)
- [TunnelsClient](https://docs.usesynth.ai/reference/sdk/tunnels-client.md)
- [Hosted Optimizers](https://docs.usesynth.ai/sdk/hosted-optimizers.md): Quickstart for hosted GEPA and GELO runs with the synth-optimizers SDK and CLI.
- [Install and Authenticate](https://docs.usesynth.ai/sdk/install-and-auth.md): Install synth-ai and configure API keys, backend URLs, and client construction.
- [In-Process Task App](https://docs.usesynth.ai/sdk/localapi/in-process.md): Run task apps entirely within your Python script with automatic tunnel management
- [Synth AI SDK Overview](https://docs.usesynth.ai/sdk/overview.md): Install synth-ai once; use SynthClient for Managed Research, Research Factory, and infrastructure surfaces.
- [Container Pools](https://docs.usesynth.ai/sdk/pools.md): Create container pools, launch rollouts, and read rollout artifacts, usage, summaries, and events.
- [SynthClient](https://docs.usesynth.ai/sdk/synth-client.md): SynthClient is the front-door Python client for research products and infrastructure surfaces.
- [Outcome Rewards](https://docs.usesynth.ai/sdk/tracing/rewards/outcome-rewards.md): Session-level rewards for filtering and evaluation
- [Tunnels](https://docs.usesynth.ai/sdk/tunnels.md): Expose local task services to hosted optimizer runs with SynthTunnel, cloudflared, or ngrok.
- [Stack Changelog](https://docs.usesynth.ai/stack/changelog.md): Public release notes for Stack channels, install paths, handoffs, and operator-facing changes.
- [Cockpit controls](https://docs.usesynth.ai/stack/cockpit.md): Keyboard controls, Agent Bridge modes, and the Stack TUI layout.
- [Configuration](https://docs.usesynth.ai/stack/configuration.md): Stack config, working directory, environments, and common environment variables.
- [Connect Synth](https://docs.usesynth.ai/stack/connect-synth.md): Sign in to unlock hosted SMR, Factories, hosted optimizers, and Synth inference from Stack.
- [Goal mode & monitor](https://docs.usesynth.ai/stack/goal-mode.md): Run long Codex goals with a sidecar monitor — curated Sidecar events instead of raw tool spam.
- [Hosted ops](https://docs.usesynth.ai/stack/hosted-ops.md): Operate hosted SMR, Research Factories, hosted optimizers, and remote sync from Stack.
- [Hosted inference](https://docs.usesynth.ai/stack/inference.md): Use Synth free aux and billed GLM inference from Stack without changing the default local worker.
- [Local-only mode](https://docs.usesynth.ai/stack/local-only.md): Run Stack locally without a Synth signup, then connect only when you want hosted features.
- [Stack MCP](https://docs.usesynth.ai/stack/mcp.md): Give Codex typed access to local optimizers and hosted Synth live ops.
- [Optimizers (GEPA)](https://docs.usesynth.ai/stack/optimizers.md): Run GEPA prompt optimization on your tasks from the Stack cockpit.
- [Stack](https://docs.usesynth.ai/stack/overview.md): Use Stack as a local operator cockpit for Synth agents, handoffs, and run evidence.
- [Quickstart](https://docs.usesynth.ai/stack/quickstart.md): Download Stack, run the local loop, and configure it with the Codex coding agent.
- [Stackd local API](https://docs.usesynth.ai/stack/stackd.md): Stack's localhost control plane — threads, events, MCP, telemetry, and exports.

## OpenAPI Specs

- [openapi](https://docs.usesynth.ai/python-sdk/api-reference/openapi.json)