Skip to content
Durable Agent Directoryby NextElevenStart with Mothership
Menu

Durable agent infrastructure · 18 listings

Agent infrastructure that survives restarts, asks before it acts, and leaves evidence.

A focused directory for teams putting agents near real systems. Every listing is marked against the same six production criteria — including what it does not claim — with links to the documentation behind each mark.

Browse the directoryStart with Mothership
  • Approval before irreversible actionsA person or policy holds the action until it is decided.
  • State that survives restartsWork resumes from the last completed step after a crash or redeploy.
  • A kill switch you controlOperators can withdraw permission so further actions do not start.
  • Evidence you can reviewWhat ran, with what inputs, and who approved it.

Flagship · Built by NextEleven

Mothership

Mothership is NextEleven’s proprietary agent infrastructure. Explore the product site and request a technical pilot conversation. Hosted seats, pricing and live access require confirmation; the public pilot page is an information and inquiry path.

By NextElevenPaid productPilot inquiry · pricing on requestUnder evaluationNot yet verified
Request a technical pilotExplore the productCriteria & limits
  • Approval gates: Not claimed.
    Approval behavior has not been verified for this directory.
  • Restart recovery: Not claimed.
    Restart and redeploy recovery have not been verified for this directory.
  • Evidence trail: Not claimed.
    Evidence recording has not been verified for this directory.
  • Kill switch: Not claimed.
    Operator stop behavior and integration coverage have not been verified for this directory.
  • Policy controls: Not claimed.
    Tool scoping and policy behavior have not been verified for this directory.
  • Sandbox isolation: Not claimed.
    No sandbox isolation is claimed.
DeadBoltContainedA local execution gate: policy and operator controls decide whether each agent action runs.HarnessDurableA Rust work assistant with exact model routing, resumable sessions, bounded workers and MCP.WitnessA provenance prototype that keeps the path from evidence to claims, with signed append-only records.

Disclosure: NextEleven runs this directory and builds the products in this band. They are listed first because they are ours. Their criteria marks follow the same rules as every other listing, and a “Verified” badge is only shown with published evidence.

Directory

Find the piece you are missing

Built inDocumented patternNot claimed
18 listings

Mothership

NextEleven · Agent workspaces & platforms
By NextElevenPaid productPilot inquiry · pricing on requestUnder evaluationNot yet verified

Proprietary agent infrastructure · technical pilot by inquiry.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Not claimed
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Proprietary

DeadBolt

NextEleven · Execution gates & kill switches
By NextElevenAgent tool / platformFree source toolLiveEvidence linked

A local execution gate: policy and operator controls decide whether each agent action runs.

  • Approval: Built in
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Built in
  • Policy: Built in
  • Isolation: Not claimed
Self-hostedOpen source

Harness

NextEleven · Agent workspaces & platforms
By NextElevenAgent tool / platformFree source toolBetaEvidence linked

A Rust work assistant with exact model routing, resumable sessions, bounded workers and MCP.

  • Approval: Not claimed
  • Recovery: Built in
  • Evidence: Not claimed
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedOpen source

Witness

NextEleven · Evidence & observability
By NextElevenAgent tool / platformFree source toolPrototypeEvidence linked

A provenance prototype that keeps the path from evidence to claims, with signed append-only records.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedOpen source

Amazon Bedrock AgentCore

Amazon Web Services · Agent workspaces & platforms
Product / platformCheck vendor pricing and termsLiveFrom public docs

Managed AWS services to run, govern and observe agents built with any framework.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Not claimed
  • Policy: Built in
  • Isolation: Built in
HostedProprietary managed service (consumption pricing)

Temporal

Temporal Technologies · Durable execution & workflows
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Durable execution platform: workflows replay from history and survive failures.

  • Approval: Documented pattern
  • Recovery: Built in
  • Evidence: Built in
  • Kill switch: Built in
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

ToolHive

Stacklok · MCP gateways & tool policy
Agent tool / platformFree source toolLiveFrom public docs

Run and manage MCP servers in isolated containers, with access policy and audit logs.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Not claimed
  • Policy: Built in
  • Isolation: Built in
Self-hostedOpen source

Docker MCP Gateway

Docker · MCP gateways & tool policy
Agent tool / platformFree source toolLiveFrom public docs

The Docker CLI plugin behind MCP Toolkit: containerized MCP servers with tool allowlists.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Not claimed
  • Policy: Documented pattern
  • Isolation: Built in
Self-hostedOpen source

LangGraph

LangChain · Agent frameworks with checkpoints
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Low-level orchestration for stateful agents, with durable execution and human-in-the-loop.

  • Approval: Built in
  • Recovery: Built in
  • Evidence: Documented pattern
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

Microsoft Agent Framework

Microsoft · Agent frameworks with checkpoints
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Build and orchestrate agents and multi-agent workflows in Python and .NET.

  • Approval: Built in
  • Recovery: Built in
  • Evidence: Documented pattern
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

Trigger.dev

Trigger.dev · Durable execution & workflows
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Open-source platform for long-running AI tasks in TypeScript, with checkpointing and waitpoints.

  • Approval: Built in
  • Recovery: Built in
  • Evidence: Documented pattern
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

agentgateway

Linux Foundation · MCP gateways & tool policy
Agent tool / platformFree source toolLiveFrom public docs

Open-source proxy for agent and tool traffic over MCP and A2A, with CEL policy.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Documented pattern
  • Kill switch: Not claimed
  • Policy: Built in
  • Isolation: Not claimed
Self-hostedOpen source

DBOS

DBOS · Durable execution & workflows
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Durable workflows and queues as a library, backed by the Postgres you already run.

  • Approval: Documented pattern
  • Recovery: Built in
  • Evidence: Documented pattern
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

E2B

E2B · Sandboxes & isolation
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Open-source infrastructure for running AI-generated code in isolated cloud sandboxes.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Not claimed
  • Kill switch: Documented pattern
  • Policy: Not claimed
  • Isolation: Built in
Self-hostedHostedOpen source

Inngest

Inngest · Durable execution & workflows
Product / platformCheck vendor pricing and termsLiveFrom public docs

Stateful step functions and AI workflows on serverless, servers or the edge.

  • Approval: Documented pattern
  • Recovery: Built in
  • Evidence: Not claimed
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedSSPL server (delayed Apache-2.0 publication); Apache-2.0 SDKs

Langfuse

Langfuse (part of ClickHouse) · Evidence & observability
Agent tool / platformFree source · check hosted pricingLiveFrom public docs

Open-source LLM engineering platform: tracing, evaluation and debugging, self-hosted or cloud.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Built in
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedOpen source

NeMo Guardrails

NVIDIA · Runtime guardrails
Agent tool / platformFree source toolLiveFrom public docs

Open-source toolkit for programmable guardrails on LLM inputs, outputs and tool calls.

  • Approval: Not claimed
  • Recovery: Not claimed
  • Evidence: Not claimed
  • Kill switch: Not claimed
  • Policy: Built in
  • Isolation: Not claimed
Self-hostedOpen source

Restate

Restate · Durable execution & workflows
Product / platformCheck vendor pricing and termsLiveFrom public docs

Lightweight durable execution with a journal-and-replay model, self-hosted or in Restate Cloud.

  • Approval: Documented pattern
  • Recovery: Built in
  • Evidence: Not claimed
  • Kill switch: Not claimed
  • Policy: Not claimed
  • Isolation: Not claimed
Self-hostedHostedBSL 1.1 (converts to Apache-2.0 four years after each release)

Browse by category

Agent workspaces & platformsWhere agents run day to day: tools, sessions, approvals and state in one place.Execution gates & kill switchesChecks at the point an action actually runs, with operator control to withdraw permission.Durable execution & workflowsEngines that persist each step so long-running work survives crashes and waits for signals.Agent frameworks with checkpointsAgent libraries with built-in persistence and human-in-the-loop interrupts.MCP gateways & tool policyProxies and runtimes that sit between agents and MCP servers to enforce access rules.Sandboxes & isolationIsolated environments for running AI-generated code and agent tools.Runtime guardrailsProgrammable rules on model and tool inputs and outputs.Evidence & observabilityTraces, records and provenance you can review after an agent acts.

14 third-party listings, last reviewed against public documentation on 2026-10-29. Something wrong or out of date? Send a correction.

Free tools, a product pilot, or a focused paid diagnosis.

Choose a tool from the directory, explore Mothership’s technical pilot path, or check fit for a software diagnosis when one issue needs investigation.

Request a technical pilotSoftware diagnosis · $250 USD