Open Managed Agents
Run Claude Managed Agents and OpenAI Agents API sessions on infrastructure you own, Apache 2.0.
Pick OpenMA when owning the runtime is a real requirement, not a slide bullet. The harness/sandbox/cloud split is genuinely modular, and an existing Claude Managed Agents or OpenAI Agents client migrates by swapping a base URL. The catch is operational: you or someone on your team has to run Node, Docker or Cloudflare, and dynamic subagent handoff still needs self-hosted Node.
Verified 1d ago · liveness 73/100 · cite: rightaichoice.com/tools/open-managed-agents
- Developers who want to own the agent loop and modify context engineering or compaction
- Teams with data-residency rules that need sessions and execution on their own infrastructure
- Existing Claude Managed Agents or OpenAI Agents API users who migrate by changing the SDK baseURL
- Products embedding an agent runtime through the TypeScript SDK, OMA CLI or Backchat
- Teams with nobody willing to operate Docker or a Cloudflare deployment
- Buyers who want a point-and-click managed agent product with no infrastructure work
- Teams that need dynamic subagent handoff but cannot run self-hosted Node
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip OpenMA if you want a point-and-click managed agent service and nobody on your team will run Docker, Node or Wrangler — the platform's value comes from owning the runtime.
BYOK means every token you spend is billed by Anthropic, OpenAI or DeepSeek directly, on top of whatever OpenMA hosting costs you.
Four ways to run it: Free at $0/mo with a $1 one-time trial credit and unlimited local runtime; Pro at $20/mo carrying $22 credit (~73 hours cloud sandbox); Max at $100/mo carrying $120 credit (~400 hours cloud sandbox); Self-Hosted at $0/mo under Apache 2.0, where you pay model providers directly via BYOK. That undercuts closed managed-agent hosting on cost, but shifts operational work to your team.
In short
Open Managed Agents — Run Claude Managed Agents and OpenAI Agents API sessions on infrastructure you own, Apache 2.0. Best for Developers who want to own the agent loop and modify context engineering or compaction, Teams with data-residency rules that need sessions and execution on their own infrastructure, Existing Claude Managed Agents or OpenAI Agents API users who migrate by changing the SDK baseURL. Free to start; paid plans from $20/mo.
What's new in Open Managed Agents
Checked 7 days agoAcross the latest 5 updates: 1 changelog entry and 4 news mentions.
How to Self-Host a Claude Tag-Style Agent
Guide that maps the Claude Tag product shape onto a self-hosted Open Managed Agents deployment, so teams can run a comparable agent without the closed host.
Claude Tag Open-Source Alternative: Self-Hosted with OpenMA
Vendor positions Open Managed Agents as the open-source, self-hosted foundation for a Claude Tag-style agent.
Migrating from Claude Managed Agents to Open Managed Agents
Migration guide covering client code, sessions, vaults and integrations for switching off Claude Managed Agents onto OpenMA.
The Architecture of Open Managed Agents: Durable Objects, R2, and the Brain/Sandbox Split
Engineering post describing Cloudflare Durable Objects with embedded SQLite as the session log, Cloudflare Containers as the sandbox, R2 for blobs, and the brain/sandbox split.
Claude Managed Agents vs Open Managed Agents: A Technical Comparison
Side-by-side comparison of API surface, runtime model, sandbox and billing between Claude Managed Agents and the open-source project.
What people actually say about Open Managed Agents — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
16 mentions across 2 sources (GitHub, Lemmy) · researched Jul 3, 2026.
Average across the 2 sources that answered — each source counts once, not each post.
- +Drop-in compatible with Claude Managed Agents HTTP API.
- +Durable sessions with strong consistency via Durable Objects + SQLite.
- +Self-host on Cloudflare Workers or Docker Compose with a five-line quickstart.
- +Apache 2.0 license allows free forking and redistribution.
- +Replaceable harness layer for custom agent loops.
- −Very early stage: only 190 stars and 20 open issues.
- −Community feedback is virtually nonexistent beyond GitHub.
- −Depends heavily on Cloudflare ecosystem, limiting portability.
- −No documented real-world production deployments yet.
- −Support relies on GitHub: no forums, chat, or dedicated help.
- • Cloudflare Workers and Durable Objects usage costs
- • LLM API key costs (BYOK)
Viability Score
How well maintained and how widely used is Open Managed Agents? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Wire-compatible with Claude Managed Agents via @anthropic-ai/sdk baseURL swap
- Wire-compatible OpenAI Agents API surface at /openai/v1
- Durable sessions with event history and recovery
- Pluggable harness layer: built-in, Pi, ACP adapter, or your own
- Pluggable sandbox runtimes: LiteBox, E2B, Daytona, BoxRun
- Sandboxed code and file execution in a separated environment
- Sandbox-free OpenAI sessions on hosted OpenMA
- Dynamic subagent handoff where children share parent files (self-hosted Node)
- MCP server connections for external tool access
- ACP adapter for connecting compatible external agents
- DeepSeek ACP Agent adapter
- TypeScript SDK for embedding OpenMA in your application
- OMA CLI for terminal access to sessions
- Backchat macOS desktop app with local and tenant sessions in one sidebar
- BYOK model keys with no token markup from OpenMA
About Open Managed Agents
Open Managed Agents (OpenMA) is an open-source runtime for AI agent sessions that speaks both the Claude Managed Agents API surface and the OpenAI Agents API surface. An app already written against @anthropic-ai/sdk or the openai SDK points its baseURL at hosted OpenMA or your own deployment and keeps working — no rewrite. The architecture splits three concerns that most agent frameworks weld together. OpenMA handles accounts, tenants, sessions and event history. A harness runs the agent loop (built-in, Pi, or an ACP adapter, or your own). A sandbox executes code and touches files (LiteBox, E2B, Daytona, BoxRun). Each is an adapter slot, so you can change context engineering or move execution somewhere else without rebuilding the platform. You drive it through Backchat, a macOS desktop app that lists local sessions and tenant sessions in one sidebar and resumes a session where it stopped, through the OMA CLI, or through the TypeScript SDK. MCP supplies tool access and ACP connects compatible external agents, including the vendor's DeepSeek ACP adapter. Self-hosting is Node plus Docker on your box, or a Cloudflare deployment; start with SQLite and add Postgres and S3 as you grow. BYOK is a design principle, not an add-on: you supply model keys and pay model costs directly, and hosted plans bill sandbox compute. Recent vendor posts position OpenMA as the open-source alternative to the closed Claude Tag and a migration target for Claude Managed Agents. It's for developers who want to own the agent loop — not buyers shopping for a point-and-click managed agent service.
Behind the Verdict
Most 'open-source agent platform' claims collapse on inspection: you get a prompt loop and a README, and everything that mattered stayed hosted. OpenMA is unusually concrete about the hard parts. Sessions with event history and recovery, a harness adapter, a sandbox adapter, a deployment target — each one swappable, each one documented. We'd reach for this when a team has outgrown a closed agent host but doesn't want to write session persistence and sandbox plumbing from scratch. The migration story is the strongest hook. If you already have code against @anthropic-ai/sdk or the openai SDK, the change is a baseURL and an environment ID, then you're pointed at app.openma.ai or your own box. The vendor has published migration guides covering client code, sessions, vaults and integrations, plus an architecture post on the Durable Objects + R2, brain/sandbox split. That's a team that expects you to actually move, not just admire. Where it bites: the feature surface depends on where you run it. Dynamic subagent handoff — children sharing parent files with their own conversations — only works on self-hosted Node, children can't spawn further children, and hosted OpenMA is the one that gives you sandbox-free OpenAI sessions. Advanced environment and plugin settings are unsupported on Node today. Read the supported-features page before you promise a customer anything. Hosted pricing is oddly shaped but honest. Free covers local runtime with integrations included. Pro at $20/mo lands $22 of wallet credit and roughly 73 hours of cloud sandbox. Max at $100/mo gives $120 credit and about 400 hours plus priority email support. You bring model keys either way, so token spend never runs through OpenMA — only compute does. Self-host stays free forever on the same Apache 2.0
Researching Open Managed Agents? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Open Managed Agents actually fits — and what changes day-one when you adopt it.
Keep the existing @anthropic-ai/sdk client, swap baseURL to the hosted OpenMA endpoint, reuse the configured model and environment ID, then create the agent and session as before.
Outcome: The app runs against OpenMA without a rewrite, and the May 12 2026 migration guide covers client code, sessions, vaults and integrations.
Deploy OpenMA on self-hosted Node and Docker starting with SQLite, connect Postgres and S3 as it grows, and attach an E2B or Daytona sandbox for code execution.
Outcome: Sessions, event history and code execution stay on your infrastructure while the agent loop and sandbox remain swappable adapters.
Use the TypeScript SDK or OMA CLI inside your own application on hosted OpenMA, run sandbox-free OpenAI sessions, and let developers pick tasks back up in Backchat on macOS.
Outcome: An agent runtime ships inside your product without building session, tenant and event-history plumbing, and model usage is billed by your provider under BYOK.
Use Cases
- Point an existing Claude Managed Agents SDK client at your own OpenMA deployment by changing the base URL.
- Point an OpenAI Agents API client at OpenMA's /openai/v1 endpoint and run sandbox-free sessions.
- Run sandboxed code execution and file operations in a runtime you choose, separate from the agent loop.
- Swap in custom context engineering or compaction by replacing the harness without rewriting the platform.
- Give agents access to your internal services through MCP servers instead of external dependencies.
- Develop and test harness modifications on Node and Docker locally before deploying to Cloudflare.
- Stand up a Claude Tag-style agent on self-hosted OpenMA rather than a closed product.
- Keep local sessions and tenant sessions side by side in Backchat and resume tasks where they stopped.
Models Under the Hood
as of 2026-10-17
Limitations
- OpenMA coordinates sessions, harnesses and sandbox adapters rather than shipping a model — you bring your own keys (BYOK), and available features depend on the harness, sandbox adapter and deployment you choose.
- The vendor's own compatibility notes state that dynamic subagent controls currently require Node, hosted OpenMA supports sandbox-free OpenAI sessions, and advanced environment and plugin settings remain unsupported; check the supported-features page before migrating.
- Subagent children cannot create further children.
- It speaks the Claude Managed Agents and OpenAI Agents API surfaces, so it assumes developers already comfortable with those SDKs.
- Backchat is macOS-only.
- Self-hosting means operating Node and Docker, or Wrangler and a Cloudflare account.
as of 2026-10-11
Verification history
We have re-verified Open Managed Agents 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Open Managed Agents tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Developers evaluating OpenMA locally who want unlimited local runtime before committing any spend.
What this tier adds
Starting tier: $0/mo with a $1 one-time trial credit, unlimited localRuntime, all integrations, and your own model keys.
Pro
$20/mo
Ideal for
Solo developer or small team running cloud sandbox sessions steadily without wanting a full self-hosted deployment.
What this tier adds
Adds $22 credit with a 10% bonus and about 73 hours of cloud sandbox, with auto-renewing wallet top-up.
Max
$100/mo
Ideal for
Teams running heavy cloud sandbox workloads who need more hours and a support escalation path.
What this tier adds
Raises the credit to $120 with a 20% bonus and about 400 hours of cloud sandbox, and adds priority email support.
Where the pricing makes sense
The company stage and team size where Open Managed Agents's pricing actually pencils out — and where peers do it cheaper.
Four ways to run it: Free at $0/mo with a $1 one-time trial credit and unlimited local runtime; Pro at $20/mo carrying $22 credit (~73 hours cloud sandbox); Max at $100/mo carrying $120 credit (~400 hours cloud sandbox); Self-Hosted at $0/mo under Apache 2.0, where you pay model providers directly via BYOK. That undercuts closed managed-agent hosting on cost, but shifts operational work to your team.
Setup time & first value
How long it actually takes to get something useful out of Open Managed Agents — broken out by persona, not the marketing-page minute.
Self-hosted on Node and Docker: an afternoon to a day for a developer comfortable with containers, following the setup guide. Cloudflare deployment adds Wrangler config, secrets and custom domains, so budget roughly a day for the first deploy. Migrating an existing Claude Managed Agents or OpenAI Agents client is minutes — it is a baseURL change plus your model key.
Switching to or from Open Managed Agents
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Claude Managed Agents: point @anthropic-ai/sdk at your OpenMA baseURL and follow the vendor's May 2026 migration guide for client code, sessions, vaults and integrations.
- →From OpenAI Agents API: change the client baseURL to /openai/v1 and run sandbox-free sessions on hosted OpenMA.
- →From LangGraph, AutoGen or CrewAI: rewrite the orchestration layer but keep session, tenant and event-history plumbing in OpenMA while you swap the harness.
- ↗To a closed managed agent host: migrate the session and vault concepts into that vendor's API surface, which removes the adapter flexibility you had.
- ↗To a framework like LangGraph or CrewAI: keep the model keys via BYOK, but rebuild orchestration because OpenMA's harness slot maps to framework graphs only by rewriting the loop.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Open Managed Agents”, and we withheld 6: 6 did not mention Open Managed Agents. We are showing none, because we could not prove any of them are about Open Managed Agents.
Official links
Featured Head-to-Head Comparisons
Open Managed Agents vs Spider Cloud
Choose Spider Cloud if you need fast, cost-effective web data extraction for AI agents or RAG pipelines, with ready-made scrapers and AI-powered extraction. Choose Open Managed Agents if you want to self-host a full agent platform compatible with Claude Managed Agents, prioritizing data control and no vendor lock-in over managed simplicity.
Open Managed Agents vs Temporal Ai
For teams building reliable, long-running AI agents that must survive failures and include human-in-the-loop, Temporal AI is unmatched with its durable execution and battle-tested ecosystem. If you need a self-hosted, wire-compatible alternative to Claude Managed Agents with full control over the agent loop and data, Open Managed Agents offers a lightweight, open-source solution at no platform cost.
Open Managed Agents vs Presto Voice
Presto Voice and Open Managed Agents serve completely different buyers. Presto Voice is a domain-specific SaaS for QSR drive-thrus delivering measurable revenue lift, while Open Managed Agents is an open-source developer tool for hosting AI agents on your own infrastructure. Choose Presto if you run a drive-thru chain and want a turnkey voice AI solution; choose Open Managed Agents if you're a developer building custom agent workflows with full control over the loop.
Popular in Agent Frameworks & Orchestration
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
Frequently Asked Questions
Best-of guides
Used Open Managed Agents? Help shape our editorial sentiment research.