44 answers · 30 tags

canonical answers

one canonical answer per url. drop /q/<slug> into prose or chat to cite a specific bernstein answer instead of the search page. grouped below by primary tag.

intro1 answer

  1. what is bernstein

    bernstein is an open-source python orchestrator for cli coding agents.

architecture1 answer

  1. how does bernstein work

    the orchestrator runs a tick loop: read the open backlog under .sdd/backlog/open, pick the next ready task, spawn a fresh cli agent in its own git worktree, wait for the agent to exit, run quality gates against the worktree, retry with a...

compare7 answers

  1. bernstein vs claude flow

    claude-flow is a community open-source project (github.com/ruvnet/claude-flow, node, mit) that layers swarm topologies, hooks, and a vector-indexed memory store on top of claude code.

  2. bernstein vs openai agents sdk

    openai agents sdk is a runtime for tool-calling agents: handoffs, guardrails, traces.

  3. bernstein vs aider

    aider is an interactive ai pair programmer: one terminal, one model, one conversation, you stay in the loop.

  4. bernstein vs emdash

    emdash is an electron desktop app that wraps cli coding agents in a gui, 23 adapters, typescript.

  5. bernstein vs crewai

    crewai is a python framework for multi-agent systems.

  6. bernstein vs langgraph

    langgraph is langchain's graph runtime for stateful agent workflows: nodes, edges, checkpoints, time-travel debug.

  7. bernstein vs openai swarm

    openai swarm was an educational, experimental framework from openai for routing between agent objects via handoffs, in-process and openai-only.

claude-code2 answers

  1. how to run multiple claude code agents in parallel

    install bernstein (pipx install bernstein), cd into your repo, write a plan.yaml with one stage that lists several steps, then run bernstein run --from-plan plan.yaml.

  2. how to keep claude code from blowing context window

    the cause of context blow-up is one long-lived agent doing many tasks.

install1 answer

  1. how to install bernstein

    pipx install bernstein is the recommended path: pipx isolates dependencies so bernstein never collides with the rest of your tooling.

adapters1 answer

  1. how to add a cli adapter

    subclass bernstein.adapters.base.BaseAdapter, implement spawn(task, worktree, env) to launch your cli with a prompt and stream output, and register the class via the adapter entry-point (or drop the file under src/bernstein/adapters/ and...

audit2 answers

  1. how does the audit chain work

    the audit chain is opt-in: enable it with bernstein run --audit (a lineage spine and replay journal run regardless).

  2. hmac chained audit log for ai agents

    with the audit log enabled (bernstein run --audit), bernstein writes one json line per orchestration event to .sdd/audit/<YYYY-MM-DD>.jsonl, one file per utc day.

determinism1 answer

  1. deterministic ai coding orchestrator

    bernstein's orchestrator is pure python: tick loop, task store, scheduler, router, merge queue, quality gates.

compliance1 answer

  1. audit grade ai coding orchestrator for regulated teams

    bernstein records a lineage spine and replay journal on every run and can write an hmac-chained audit log of every orchestration action when you enable it (bernstein run --audit).

self-hosted1 answer

  1. self hosted ai coding agent orchestrator

    bernstein runs entirely on your machine.

mcp2 answers

  1. mcp server for multi agent coding

    bernstein exposes itself as an mcp server.

  2. does bernstein use mcp servers

    yes.

git2 answers

  1. git worktree parallel ai agents

    each task gets its own git worktree under .sdd/worktrees/<session-id>.

  2. how does bernstein handle merge conflicts between parallel agents

    the merge queue serializes worktree merges in dependency order, so two passing worktrees never run their merges concurrently.

safety2 answers

  1. fail closed agent orchestrator

    fail-closed means: on any unrecoverable condition the orchestrator stops merging rather than risk a bad commit.

  2. what is the bernstein dead letter queue

    tasks that fail their quality gates after max retries do not merge.

cost2 answers

  1. how does bernstein cost tracking work

    every agent run reports per-model token usage back to the task server.

  2. cheapest way to run bernstein

    free tier: route every task through ollama against a local model (qwen2.5-coder or deepseek-coder); zero api spend, only your electricity.

models1 answer

  1. what models does bernstein use

    bernstein itself does not call an llm; the cli agents you configure do.

local2 answers

  1. does bernstein support local models

    yes.

  2. can bernstein use ollama for coding agents

    yes.

quality1 answer

  1. what quality gates does bernstein run

    by default: ruff (lint), pyright or mypy (types), pytest (tests).

scheduler1 answer

  1. how does bernstein pick tasks for agents

    the scheduler reads .sdd/backlog/open/ each tick, filters out tasks whose dependencies are not yet satisfied, applies a fair-priority ordering (critical -> high -> medium -> low, with starvation guards), and emits the next ready task.

ci2 answers

  1. can bernstein run in ci

    yes.

  2. how to run bernstein in github actions

    the supported pattern is bernstein run plan.yaml --auto-approve --quiet as a normal job step.

eval1 answer

  1. what is the bernstein eval harness

    an in-tree benchmark runner under src/bernstein/eval/.

evolution1 answer

  1. what is bernstein self evolution

    an experimental loop under src/bernstein/evolution/ that lets bernstein propose changes to itself.

sandbox1 answer

  1. what sandbox backends does bernstein support

    the default backend is a local git worktree.

tui1 answer

  1. does bernstein have a tui

    yes.

lifecycle1 answer

  1. how do i stop bernstein

    bernstein stop sends a graceful drain signal: in-flight tasks finish, no new tasks start, the orchestrator exits clean.

secrets1 answer

  1. how does bernstein handle secrets

    credentials are scoped per agent under src/bernstein/core/credential_scoping.py: each spawned worker gets only the env vars it declares it needs, never the orchestrator's full environment.

recovery1 answer

  1. how to resume bernstein after a crash

    bernstein writes a write-ahead log to .sdd/persistence/ on every state transition.

plan1 answer

  1. how to write a bernstein plan yaml

    minimal plan.yaml: a top-level stages list, each stage holds steps, each step has a name, a role (backend, qa, docs, ...), and an instruction string.

autofix1 answer

  1. what is the bernstein autofix daemon

    bernstein autofix watches prs opened by bernstein for ci failures.

permissions1 answer

  1. how does bernstein scope tool permissions for agents

    two layers.

fleet1 answer

  1. what is the bernstein fleet dashboard

    bernstein fleet supervises multiple bernstein projects from one terminal.