Bernstein is live · free plan, paid from $25/mo · last verified 21 Oct 2026 · viability healthy
Bernstein
Bernstein runs CLI coding agents in parallel with deterministic scheduling and a verifiable audit chain.
Most multi-agent orchestrators ask you to trust a dashboard; Bernstein's hash chain recomputes, which is a materially different claim. Keeping the model out of the scheduling tick is what makes replay and offline verification mean anything — drop an LLM in there and the run stops reproducing. The honest caveats: single maintainer, beta status, terminal-first, and a routing saving the vendor's own cost page calls a heuristic that collapses toward zero on flaky-test repos.
Verdict generatedVerified 8d ago · liveness 80/100 · cite: rightaichoice.com/tools/bernstein
- Engineering teams running several CLI coding agents in parallel who only want verified diffs reaching main
- Compliance-driven teams in finance or healthcare that must show evidence of what agents did and when
- Platform engineers needing reproducible, replayable multi-agent runs on their own infrastructure
- Forward-deployed engineers dropping into client repos with tight test suites and mechanical work
- Teams wanting a no-code, GUI-first agent builder rather than terminal software
- Anyone needing a managed service with a vendor support contract or SLA
- Beginners without CLI and git worktree fluency
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 need a managed service with a support contract behind it, or if your repo's tests are too flaky for cheapest-passing-model routing to clear your threshold.
The adapters are the real bill: Bernstein drives Claude Code, Codex, Cursor and Gemini CLI that you already pay for, so a five-agent parallel run multiplies your existing LLM spend rather than replacing it.
The runtime is free under Apache 2.0 at any team size, which undercuts seat-priced orchestrators outright. The $25/mo GitHub Sponsors tier is optional upkeep for the maintainer, documented as roughly 4% of a $600/mo combined LLM bill. Your true cost sits in the CLI agents Bernstein drives — the calculator suggests $180–$360/mo shifted on $600/mo, but that band is heuristic, not contracted.
In short
Bernstein — Bernstein runs CLI coding agents in parallel with deterministic scheduling and a verifiable audit chain. Best for Engineering teams running several CLI coding agents in parallel who only want verified diffs reaching main, Compliance-driven teams in finance or healthcare that must show evidence of what agents did and when, Platform engineers needing reproducible, replayable multi-agent runs on their own infrastructure. Free to start; paid plans from $25/mo.
What is Bernstein?
Bernstein — Bernstein runs CLI coding agents in parallel with deterministic scheduling and a verifiable audit chain. Best for Engineering teams running several CLI coding agents in parallel who only want verified diffs reaching main, Compliance-driven teams in finance or healthcare that must show evidence of what agents did and when, Platform engineers needing reproducible, replayable multi-agent runs on their own infrastructure. Free to start; paid plans from $25/mo.
Bernstein is an open-source governance layer for AI coding agents. You hand it a goal, a deterministic Python scheduler breaks it into tasks, and it spawns CLI agent adapters in parallel — 53 of them at v3.22.0, spanning Claude Code, Codex, Gemini CLI, Cursor, Aider, GitHub Copilot and more — each isolated in its own git worktree. A janitor stage runs lint, types and tests against every diff, with optional cross-model review, and only what passes lands on main. The design choice everything hangs on: no model sits inside the coordination tick. Who runs, who's blocked, what merges — all handled deterministically, so the same plan replays byte-identically. Model selection is separate and opt-in, a capability router plus an epsilon-greedy contextual bandit that prefers the cheapest model whose recent pass rate on that task type clears your threshold. Tasks that declare an artifact contract close with signed deliverables instead: research dossiers, datasets, audit packs. Evidence is produced rather than screenshotted. Run with --audit and Bernstein writes an HMAC-chained event log under .sdd/audit/, one JSONL file per UTC day, each entry referencing the previous hash. Configure a signing key and the run also emits an Ed25519-signed receipt that bernstein verify receipt checks offline against a pinned public key — no HMAC key, no .sdd/ needed. Telemetry defaults to local-only; you can push an audit-grade trail to your own OTel collector, Datadog, Splunk or S3. It sits on your laptop, on-prem behind a firewall, air-gapped, on Kubernetes, or experimentally on Cloudflare Workers running in your own account. Apache 2.0, beta, built by one maintainer — no managed hosting, no GUI-first builder, no vendor support org. Think of it as the governance and verification layer around the CLIs your team already runs, not a replacement for Claude Code or Codex.
Key features
- Deterministic Python scheduler with no model in the coordination loop, replaying plans byte-identically
- 53 CLI agent adapters including Claude Code, Codex, Gemini CLI, Cursor, Aider, GitHub Copilot and Muse Code
- Parallel agent execution with each task isolated in its own git worktree
- Janitor verification stage gating lint, types, tests and PII per diff before merge
- Optional cross-model review as an extra janitor check
- Artifact contract mode closing research dossiers, datasets and audit packs with signed lineage receipts
- HMAC-chained audit log under .sdd/audit/, one JSONL file per UTC day linked by previous hash
- Ed25519-signed run receipts verified offline with bernstein verify receipt, no HMAC key required
- bernstein audit verify recomputes signature, parent-hash chain and content hash without a rerun
- Epsilon-greedy contextual bandit routing each task type to the cheapest model that passes your tests
- Opt-in audit-grade telemetry pushed to your own OTel collector, Datadog, Splunk or S3
- Sandbox execution modes: local Docker, Cloudflare Sandbox, Playwright sandbox for UI agents
- Two operator surfaces over one task API: bernstein live terminal dashboard and bernstein gui serve
- Self-hosted deployment on laptop, on-prem, Kubernetes or Cloudflare Workers (experimental)
- Agent Plugins v1.0.0 manifests and bernstein audit diagnose for run inspection
Behind the verdict
Where Bernstein earns its keep is the moment you have more than one CLI agent in play and no way to prove what they did. The verified-diff gate and the audit chain answer a question compliance teams actually ask, and they answer it with something a reviewer can check on their own machine. We'd reach for it on repos with tight tests and a lot of mechanical work — typed refactors, test scaffolding, lint sweeps — because that's precisely where routing to cheaper models pays off and where the merge gate has real signal. Pick it when you're already running Claude Code or Codex by hand and want parallel execution without trusting agent output blind. Pick it harder when someone in the room needs an offline-verifiable record of which agent touched which file and when. The Ed25519 receipt path is the part competitors rarely bother building. Pass on it if you want a no-code agent builder or a hosted service with an SLA. Nothing here is managed; you provide the machine, the repo and the tests. Pass too if your team isn't comfortable in a terminal with git worktrees — the operator surfaces are a live dashboard and a browser GUI over one task API, not a point-and-click flow. The blunt caveat sits in the vendor's own cost calculator: the 40–80% routable fraction is a heuristic, not a measured number. On flaky tests or thin coverage, the bandit falls back to premium models and the saving shrinks toward nothing. Test quality is load-bearing here in a way it isn't for a simpler orchestrator. Compared with Claude Code's own sub-agents, Bernstein spreads the same idea across 53 adapters and verifies output against your tests rather than trusting it — and Claude Code remains the most common backend inside it, so the two aren't rivals. Closest genuine alternatives are hand-rolled
Models under the hood
as of 2026-10-20
Limitations
- Bernstein is beta software built and maintained by one developer as a self-bootstrapped solo project with no investor money, so the bus factor is one.
- It is a governance/orchestration layer rather than a model — it drives 53 CLI coding agents (Claude Code, Codex, Cursor, Aider, Gemini CLI and others) that you supply and pay for yourself.
- The cost figures on the pricing page are explicitly labelled a heuristic, not a promise; real savings depend on whether cheaper models pass your project's tests, and on flaky or low-coverage repos the benefit shrinks toward zero.
- Some verification paths need the install's audit key: replaying the per-line HMAC chain requires the key the chain was written with, which by design lives outside the audit volume — export an hmac-chain+pubkey pack if you need a reviewer without the key to authenticate.
- The Cloudflare Workers runtime is documented as experimental.
as of 2026-10-17
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.
Is Bernstein still active?
Bernstein is live · free plan, paid from $25/mo · last verified 21 Oct 2026 · viability healthy.
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: October 2026
How we score →What's new in Bernstein
Checked 5 days agoAcross the latest 2 updates: 2 changelog entries.
Bernstein 3.14.159 tightens plan validation, removes five CLI commands
Version 3.14.159 makes the plan loader raise named errors on malformed fields, cuts five commands with deprecation windows, and CI-gates reference docs against the registered CLI. Adds Muse Code adapter, Agent Plugins v1.0.0 manifests, offline-verifiable signed run receipts, and
Bernstein 3.x recap covers deterministic replay and evidence bundles
Retrospective on 2.0 through 3.5.0: signed lineage spine, replay by default, content-addressed evidence bundles, cross-machine ledger resume, detached runs, tournament selection, cost-aware dispatch under USD caps, SPIFFE identity, and an MCP Tasks surface.
What do users say about Bernstein?
What people actually say about Bernstein — is it worth it?
Our community sentiment research written Jul 3, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
How we verified this
Every listed fact about Bernstein— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 21 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified Bernstein 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.
- — changed who should skip it, who it suits, pricing tiers, integrations, features, summary, description, our verdict
- — changed pricing page
- — changed our verdict, who should skip it, features, use cases, setup time, pricing fit, integrations, description, limitations, pricing tiers, summary, switching in, switching out, hidden costs
- — changed pricing page, models, limitations
- — changed who it suits, summary, who should skip it, our verdict, features, description, pricing tiers, SSL certificate
- — changed limitations, platforms, models, pricing page, pricing model
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does Bernstein cost in 2026?
Bernstein: Free to start; paid plans from $25/mo. Pricing last verified 21 Oct 2026.
Plans compared
Every published Bernstein tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Open Source | Free | — | Any team that can run beta terminal software and wants deterministic, auditable multi-agent runs without a licence fee.Adds: Starting tier: full Apache 2.0 install via pipx, brew, uv or docker with the deterministic scheduler, 53 CLI adapters, janitor gating, audit chain and on-prem or air-gapped deployment. |
| GitHub Sponsors | $25/mo | monthly | Teams already getting value from Bernstein that want to fund a solo maintainer, at roughly 4% of a $600/mo combined LLM bill.Adds: Adds optional support for the maintainer only — routing and model selection stay on-prem and model-agnostic, so behaviour is identical to the free tier. |
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.
Where the pricing makes sense
The company stage and team size where Bernstein's pricing actually pencils out — and where peers do it cheaper.
The runtime is free under Apache 2.0 at any team size, which undercuts seat-priced orchestrators outright. The $25/mo GitHub Sponsors tier is optional upkeep for the maintainer, documented as roughly 4% of a $600/mo combined LLM bill. Your true cost sits in the CLI agents Bernstein drives — the calculator suggests $180–$360/mo shifted on $600/mo, but that band is heuristic, not contracted.
Has Bernstein's pricing changed?
No plan price for Bernstein has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick Bernstein?
Pick Bernstein if you are Engineering teams running several CLI coding agents in parallel who only want verified diffs reaching main, Compliance-driven teams in finance or healthcare that must show evidence of what agents did and when, Platform engineers needing reproducible, replayable multi-agent runs on their own infrastructure, Forward-deployed engineers dropping into client repos with tight test suites and mechanical work.
Look elsewhere if you are Teams wanting a no-code, GUI-first agent builder rather than terminal software, Anyone needing a managed service with a vendor support contract or SLA, Beginners without CLI and git worktree fluency, Repos with flaky tests or thin coverage, where cheapest-passing routing falls back to premium models.
Skip it if Skip Bernstein if you need a managed service with a support contract behind it, or if your repo's tests are too flaky for cheapest-passing-model routing to clear your threshold.
GeneratedReal-world workflow fit
Concrete scenarios for the personas Bernstein actually fits — and what changes day-one when you adopt it.
Install with pipx, run bernstein init and bernstein.yaml, then fire one goal across Claude Code and Codex in parallel git worktrees with BERNSTEIN_AUDIT=1 set so every tick lands in .sdd/audit/2026-07-24.jsonl.
Outcome: You review only the diffs the janitor passed on lint, types and tests, and you keep a hash chain you can hand to security without rerunning anything.
Land in a client codebase with tight tests, scope each agent's credentials through the secrets broker, and let the bandit route mechanical lint fixes and test scaffolding to cheaper models while architecture tasks stay on premium.
Outcome: An AI crew stands up without provisioning a server, the client's keys stay out of the environment, and the run report's cost column shows what routing actually saved.
Declare an artifact contract on research and audit-evidence tasks, sign with a key, then export a pack with --signature-kind hmac-chain+pubkey for a reviewer who holds only the public key.
Outcome: The reviewer authenticates the deliverable and the chain head offline, with no HMAC key and no .sdd/ directory on their machine.
Use cases
- Run several CLI coding agents in parallel on one repo and merge only the diffs that pass tests, lint and type checks.
- Hand a reviewer who never ran the system an audit chain they can verify offline from the artifacts alone.
- Route each task type to the cheapest model whose recent pass rate clears your threshold, using bandit arms.
- Stand up an AI engineering crew on a client repo with per-agent credential scoping and no server to provision.
- Deploy a headless multi-agent pipeline on an air-gapped or firewall-bound network.
- Close non-code work — research dossiers, datasets, audit evidence packs — on artifact contracts with signed lineage receipts.
- Compare CLI coding agents on identical tasks using reproducible logs from the same plan.
- Run the swarm on Cloudflare Workers with R2 and D1 holding workspace and metering state, on your own account.
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.
Install is one command (pipx, brew, uv or docker), then bernstein init — most engineers reach a first orchestrated task in under an hour. Budget longer for the forward-deployed and compliance paths: the bandit needs a pass-rate history before routing is meaningful, and audit-key handling plus artifact contracts take an afternoon to wire.
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 hand-run CLI agents: keep Claude Code, Codex or Cursor installed, add bernstein.yaml and let the scheduler dispatch the same CLIs in isolated worktrees.
- →From a dashboard orchestrator: drop the shared workspace model, point Bernstein at your repo, and let janitor gating replace manual diff review.
- →From ad-hoc shell scripts: express the loop as a goal plus policy as code so the plan becomes replayable instead of imperative.
- →From a hosted SaaS runner: install on-prem or air-gapped and reproduce runs locally — nothing forces a SaaS hop.
- ↗To a managed agent platform: export the audit journal and receipt files first; the HMAC chain only replays with the original install's audit key.
- ↗To manual CLI use: keep the git history and artifact receipts; the deterministic scheduler and bandit routing history do not travel.
- ↗To a different orchestrator: re-declare policy as code in the new tool's format, since bernstein.yaml has no standard equivalent.
Integrations
Resources & guides
- Documentationbernstein.run
Docs · Bernstein
Full product docs from bernstein.run
- Guidebernstein.run
Guides · Bernstein
In-depth how-to from bernstein.run
- Documentationbernstein.run
Operations · Bernstein
Full product docs from bernstein.run
- API Referencebernstein.run
Reference · Bernstein
Methods, params, types from bernstein.run
- Documentationbernstein.run
Architecture · Bernstein
Full product docs from bernstein.run
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
What are the best Bernstein alternatives?
The closest alternatives to Bernstein we list are Shep, Lanes and Rover. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Shep | Free | Free | — |
| Lanes | Freemium | Pricing not published | — |
| Rover | Free | Free | — |
Alternatives to Bernstein
View all
Shep
Open-source, local-first orchestrator that runs parallel AI coding agents in isolated git worktrees, from a one-line idea to a merged pull request.
Tools that pair well with Bernstein
Common stack mates teams adopt alongside Bernstein, with the specific reason each pairing earns its keep.
Shep
Open-source, local-first orchestrator that runs parallel AI coding agents in isolated git worktrees, from a one-line idea to a merged pull request.
Lanes
macOS desktop app that runs parallel AI coding agents on an issue board, each in its own git worktree, with a self-hostable MCP endpoint for governed access.
Rover
Rover runs multiple AI coding agents in parallel, each in its own isolated sandbox on your local machine.
Bernstein vs its closest rival
Bernstein's closest head-to-head is with 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. Verdict generated
Head-to-head comparisons
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.
GeneratedBernstein 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.
GeneratedBernstein 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.
GeneratedFrequently asked questions
Categories
Best-of guides
Used Bernstein? Help shape our editorial sentiment research.