Bernstein
Open-source governance layer for AI agents — deterministic Python orchestrator with tamper-proof, replayable runs.
Bernstein delivers rare determinism and auditability for multi-agent coding, with a zero-LLM scheduler and HMAC chain that actually recomputes. It’s the right call for compliance-driven teams that need provable runs, but the CLI-first, self-hosted nature will alienate anyone wanting a managed GUI. If you can stomach the setup, it’s the most trustworthy open-source option today.
Verified 14d ago · liveness 68/100 · cite: rightaichoice.com/tools/bernstein
- Developers running multiple AI coding agents in parallel with verification
- Teams needing audit-ready, replayable orchestration for compliance
- Organizations in finance and healthcare that must prove agent actions
- Solo developers automating complex multi-step coding tasks
- Users wanting a GUI-first no-code AI agent builder
- Teams needing managed cloud hosting (Bernstein is self-hosted)
- Beginners unfamiliar with CLI tools and git worktrees
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 Bernstein if you want a managed cloud GUI, a no-code visual builder, or real-time streaming chat—it's a CLI-first, self-hosted orchestrator that demands comfort with git worktrees and YAML.
You must supply and manage your own coding agents (Claude Code, Codex, etc.) and their API costs—Bernstein itself is free but the underlying LLM usage isn't.
Bernstein is free (Apache 2.0) and self-hosted, so the only costs are your own infrastructure and LLM API usage. Compared to managed SaaS like OpenAI Codex or GitHub Copilot Workspace, which charge per-seat or per-usage, Bernstein can be cheaper if you already have the infra, but it requires more technical effort to run.
In short
Bernstein — Open-source governance layer for AI agents — deterministic Python orchestrator with tamper-proof, replayable runs. Best for Developers running multiple AI coding agents in parallel with verification, Teams needing audit-ready, replayable orchestration for compliance, Organizations in finance and healthcare that must prove agent actions. Free to use.
What's new in Bernstein
Checked 6 days agoAcross the latest 1 update: 1 changelog entry.
What people actually say about Bernstein — is it worth it?
We scanned public community sources for Bernstein on Jul 3, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Our own analysis of that scan says the posts were off-subject. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
Viability Score
How well maintained and how widely used is Bernstein? 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: September 2026
How we score →Key Features
- Deterministic Python scheduler with zero LLM in coordination loop
- Runs 40+ CLI coding agents in parallel (Claude Code, Codex, etc.)
- Git worktree isolation for each agent
- Built-in verification: lint, type checks, tests gate every merge
- HMAC-chained audit log, one JSONL per UTC day, per-entry hash references
- Ed25519-signed run receipts binding journal and lineage heads
- Replayable runs with always-on replay journal
- Detached runs for background execution
- Tournament selection with no model in decision path
- Cost-aware dispatch and cost guards
- Secrets broker for credential management
- Playwright sandbox for UI agents
- Plugin system (Pluggy) for custom adapters and hooks
- MCP server/client for tool interoperability
- A2A capability cards for agent-to-agent communication
About Bernstein
Bernstein is the open-source governance layer for AI agents. Instead of trusting one agent’s output, it runs 40+ CLI coding agents in parallel, each in its own git worktree, and merges only verified diffs. The core scheduler is pure Python—no LLM in the coordination loop—so the same plan replays byte-identically from the always-on replay journal. Built by one maintainer, it’s beta software with a growing community: 91 contributors and 2,998 closed PRs to date. Every step is verifiable. Flip on `--audit` and Bernstein writes an HMAC-chained event log, one JSONL file per UTC day, where each entry references the previous hash. Tampering breaks the chain. A reviewer who never ran the system can check the record offline with `bernstein audit verify`—no rerun needed. Configure a signing key, and runs also produce an Ed25519-signed receipt that binds journal and lineage heads; verify that alone with `bernstein verify receipt`. It’s a recomputable hash chain, not a screenshot. Beyond code, Bernstein handles non-code work like research, datasets, and audit evidence packs, closing on artifact contracts with signed lineage receipts. The latest release, v3.14.159, adds a Muse Code adapter, Agent Plugins v1.0.0 manifests, offline-verifiable signed run receipts, and `bernstein audit diagnose`. Detached runs let agents work in the background; cost-aware dispatch and guards keep spend in check; a secrets broker manages credentials; and a Playwright sandbox runs UI agents. It integrates with Claude Code, Codex, Gemini CLI, Aider, and more. Bernstein runs on your laptop, on-prem, air-gapped, on Cloudflare Workers (experimental), or as a Kubernetes cluster. It’s Apache 2.0 licensed with opt-in telemetry that defaults to local-only. If you need reproducible, auditable multi-agent coding on your own infrastructure, Bernstein is the deterministic choice—not a SaaS hop. Compared to cloud offerings like OpenAI Codex or GitHub Copilot Workspace, Bernstein keeps everything on your
Behind the Verdict
We’ve tested a lot of agent orchestrators, and Bernstein is the first one that doesn’t ask you to trust it. The zero-LLM scheduler is the core trick—because no model decides who runs or what merges, the same plan replays byte-identically. That’s not a feature you can fake, and it’s the foundation for everything else. The HMAC audit chain is the headline. Every step is signed in order, on disk, and tampering breaks verification. A reviewer who didn’t run the system can check the record offline with `bernstein audit verify`—no rerunning. The Ed25519 receipt goes further: verify it alone, no key, no .sdd/ directory. For compliance-heavy shops, that’s the difference between an audit that’s credible and one that’s a hand-wave. Where Bernstein shines is multi-agent orchestration across 40+ CLI adapters. Claude Code, Codex, Gemini CLI, Aider—you name it. Each agent runs in its own git worktree, and lint, types, and tests gate every merge. It’s not just parallel execution; it’s verified parallel execution. If you’re already using Claude Code, Bernstein treats it as a backend, not a replacement. That’s a smart pivot. The release cadence is aggressive. v3.14.159 landed in August 2026 with named plan-loader errors, a Muse Code adapter, Agent Plugins v1.0.0, and offline-verifiable receipts. The maintainer is clearly shipping fast, but that also means you’re riding a moving target. Read the changelog before upgrading. Where it bites: setup is not trivial. The stranger’s ninety minutes blog post is honest—one first-use report described ninety minutes of setup that ended in nothing. The CLI-first approach and git worktree isolation assume you’re comfortable with the terminal. If you want a GUI, look elsewhere. Also, the scheduler is deterministic, but model selection and
Researching Bernstein? 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 Bernstein actually fits — and what changes day-one when you adopt it.
You install Bernstein and run 'bernstein -g "Add JWT auth with refresh tokens, tests, and API docs"' on your repo.
Outcome: Within minutes, Bernstein decomposes the goal, spawns agents in parallel worktrees, runs lint/tests, and merges passing diffs—you get a clean git history and working code.
You parachute onto a client repo, configure per-agent credential scoping, and launch a multi-agent crew to implement a module.
Outcome: Agents work in isolated worktrees, every step is HMAC-signed, and you hand the client an auditable, replayable record for compliance review.
You enable --audit and configure a signing key, then run a multi-agent coding task in an air-gapped environment.
Outcome: A third-party reviewer can verify the signed receipt offline without rerunning the system, proving every step is untampered.
Use Cases
- Orchestrate multiple AI coding agents to implement features in parallel.
- Automate code review and merge only passing, audited changes.
- Run a headless multi-agent CI/CD pipeline with audit trails.
- Deploy a compliant, air-gapped coding assistant in regulated environments.
- Evaluate and compare coding agents on the same tasks with reproducible logs.
- Forward-deploy an AI engineering crew onto client repos with per-agent credential scoping.
Models Under the Hood
as of 2026-09-13
Limitations
- Bernstein is beta software, built and maintained by a single developer (Alex Chernysh) as a self-bootstrapped solo project, so some rough edges are expected.
- It is a deterministic Python scheduler that coordinates CLI coding agents in parallel with no LLM in the coordination loop, and it runs on your laptop, on-prem, or air-gapped.
- State lives locally in a .sdd/ directory, and you must supply and manage your own underlying coding agents and infrastructure.
as of 2026-08-26
Verification history
We have re-verified Bernstein 7 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 7 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 Bernstein tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0/mo
Ideal for
Solo developers and teams who want a free, self-hosted orchestrator and are comfortable with CLI and git.
What this tier adds
Starting tier: $0/mo, includes all core features like deterministic scheduling, 48 adapters, and HMAC audit log.
Where the pricing makes sense
The company stage and team size where Bernstein's pricing actually pencils out — and where peers do it cheaper.
Bernstein is free (Apache 2.0) and self-hosted, so the only costs are your own infrastructure and LLM API usage. Compared to managed SaaS like OpenAI Codex or GitHub Copilot Workspace, which charge per-seat or per-usage, Bernstein can be cheaper if you already have the infra, but it requires more technical effort to run.
Setup time & first value
How long it actually takes to get something useful out of Bernstein — broken out by persona, not the marketing-page minute.
For a solo developer: install via pipx or brew and run 'bernstein init'—most get their first orchestrated task done within an hour. For a forward-deployed engineer: expect 2-3 hours to configure per-agent credentials and adapters. For a compliance team: plan an afternoon to set up the HMAC key, air-gap configuration, and verification workflow.
Switching to or from Bernstein
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From OpenAI Codex CLI: replace your single-agent runs with Bernstein's parallel orchestration; you can keep using Codex as an adapter.
- →From GitHub Copilot Workspace: adopt Bernstein's CLI-first approach and use the Copilot adapter to preserve your existing agent.
- ↗To OpenAI Codex or GitHub Copilot Workspace: if you decide you need a managed cloud GUI, you'll lose the audit trail and determinism; your .sdd/ state won't transfer directly.
- ↗To Kubernetes-native CI: you can export Bernstein's lineage and verify logs as evidence, but you'll need to reimplement the workflow.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Bernstein”, and we withheld 6: 6 could not be judged, because “Bernstein” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Bernstein.
Official links
Tools that pair well with Bernstein
Common stack mates teams adopt alongside Bernstein, with the specific reason each pairing earns its keep.
Zhipu GLM
Zhipu GLM delivers open-source LLM models, MaaS APIs, and autonomous agents for Chinese enterprises and developers.
OpenHands
Open-source platform for autonomous cloud coding agents that fix bugs, review PRs, and automate workflows.
OpenAI Agents SDK
OpenAI Agents SDK: Lightweight Python framework for building multi-agent workflows with handoffs, sandboxing, and voice.
Featured Head-to-Head Comparisons
Bernstein vs Presto Voice
Presto Voice and Bernstein serve entirely different domains. Choose Presto Voice if you run a QSR chain and want to automate drive-thru orders with proven upselling ROI. Choose Bernstein if you are a developer needing auditable parallel coding agent orchestration at zero cost. Evaluate based on your role, not feature comparison.
Bernstein vs Locus Robotics
Locus Robotics and Bernstein serve completely different domains: one automates physical warehouse workflows, the other orchestrates AI coding agents. If you're a warehouse manager seeking flexible AMR automation, Locus Robotics with its new Locus Array and RaaS model is the clear choice. If you're a developer needing audit-grade, deterministic multi-agent coding orchestration, Bernstein's open-source and free features—now with a web UI in v2.0—are unmatched. Choose based on your primary problem: physical logistics vs. software development.
Bernstein vs Truleo
Truleo is purpose-built for law enforcement agencies needing to surface leads from siloed data with automated jail call and BWC analysis. Bernstein is a free, open-source orchestrator for developers running parallel AI coding agents with audit-grade logging. Choose based on domain: law enforcement or software development.
Alternatives to Bernstein
View all
Zhipu GLM
Zhipu GLM delivers open-source LLM models, MaaS APIs, and autonomous agents for Chinese enterprises and developers.
OpenHands
Open-source platform for autonomous cloud coding agents that fix bugs, review PRs, and automate workflows.
OpenAI Agents SDK
OpenAI Agents SDK: Lightweight Python framework for building multi-agent workflows with handoffs, sandboxing, and voice.
Frequently Asked Questions
Categories
Best-of guides
Used Bernstein? Help shape our editorial sentiment research.