Feature Matrix

May 21, 2026 ยท View on GitHub

Shipped capabilities in Bernstein, verified against src/bernstein/.

The "Docs status" column reflects whether the page-level reference exists (Full) or whether the capability is documented in source / module docstrings only (Brief).


Core orchestration

CapabilityDocs statusNotes
Goal-based run (-g)FullMain entry flow
Seed-file run (bernstein.yaml)FullAuto-discovery supported
Plan-file execution (stages/steps)Fullbernstein run plan.yaml
Retry + escalation plumbingFullIn task lifecycle, with configurable retries
Completion verification (janitor + signals)FullAPI + getting started coverage
Process-aware stop/drainFullGraceful and force stop, drain mode
Multi-cell orchestrationBriefImplemented in multi_cell.py
Fast-path executionBriefTrivial tasks skip LLM agent entirely (fast_path.py)
Plan mode (human approval)Full--plan-only, --from-plan, approval routes
Headless modeFull--headless for CI/overnight
Dry-run modeFull--dry-run previews plan without spawning

State and persistence

CapabilityDocs statusNotes
File-based state in .sdd/FullPrimary operating model
Metrics/trace persistenceFullPaths documented, JSONL schema
Lessons/memory persistenceBriefStored and injected
Storage backends (memory/postgres/redis)FullConfig + doctor coverage
Session persistence (fast resume)Briefsession.py - resume after stop/restart
Bulletin board (cross-agent messaging)BriefAppend-only, used by agents for handoff

Observability

CapabilityDocs statusNotes
/status and task APIFullCore API documented
Prometheus /metricsBriefEndpoint is real; Grafana dashboards are user-defined
OTLP telemetry initializationBriefWiring exists in core/observability/
Retrospective reporting (retro)FullCLI coverage present
Cost analysis (cost, history/anomaly hooks)Fullbernstein cost, cost anomaly detection active
Per-agent token progressBriefTracked in api_usage.py, surfaced in bernstein status
Session analyticsBriefbernstein recap shows session-level stats
Agent activity trackingBriefActivity metrics in metrics/
Debug bundleBriefbernstein debug, collects logs/state/config for triage

Safety and governance

CapabilityDocs statusNotes
Quality gates (lint, type-check, tests)FullPresent in run flow; extended with coverage, benchmark, arch conformance, mutation testing gates
PII scan quality gateBriefActive, auto-installed via log_redact.py
Rule enforcement (.bernstein/rules.yaml)FullEnforcement behavior documented
Log redaction (PII filter)BriefActive
Audit and verification commandsBriefbernstein audit seal/verify, Merkle proofs
HMAC-chained audit logBriefTamper-evident, daily rotation
Execution WALBriefHash-chained, crash recovery, determinism fingerprinting
Circuit breakerFullHalts misbehaving agents, writes SHUTDOWN signal
Token growth monitorBriefAuto-intervention on runaway consumption
Cost anomaly detectionBriefZ-score based, acts via task completion
Peak-hour schedulingBriefpeak_hour_router.py - cost-aware time-of-day routing
Agent loop detectionBriefKills agents in edit-loop cycles
Deadlock detectionBriefWait-for graph, automatic victim selection
Cross-model verificationBriefDifferent model reviews completed diffs (opt-in)
Behaviour anomaly detectionBriefcore/observability/behavior_anomaly.py - flags agents whose runtime metrics deviate statistically from baseline
Agent run manifestBriefHashable workflow spec for SOC2 evidence
Context degradation detectorBriefMonitors quality over time, restarts when degraded
Progressive permission promptsBriefPer-agent permission levels

Ecosystem and integrations

CapabilityDocs statusNotes
Agent catalog/discoveryFullbernstein agents sync/list/discover/match/showcase (43 CLI agent adapters)
GitHub App and CI fix flowsFullbernstein ci fix <url>, github setup
Trigger sources (github, slack, file_watch, webhook)BriefSource adapters available
Plugin hooks (pluggy)FullSDK docs in CONTRIBUTING.md
Cluster/worker primitivesFullbernstein worker --server URL, cluster routes documented
Multi-repo workspacesFullworkspace: in bernstein.yaml, workspace CLI
MCP server modeBriefbernstein mcp, MCP server in mcp/server.py
MCP tool registryBriefAuto-discovery and per-task config
MCP catalog clientBriefbernstein mcp catalog browse/search/install - installable server catalog (core/protocols/mcp_catalog/)
ACP native bridgeFullbernstein acp serve --stdio|--http :PORT - IDE-native bridge (core/protocols/acp/); see reference/acp-bridge.md
Protocol negotiationBriefprotocol_negotiation.py - runtime protocol version handshake
Schema registryBriefschema_registry.py - versioned message schemas for protocols
Credential vaultBriefbernstein connect <provider>, bernstein creds list/revoke/test - OS-keychain token storage (core/security/vault/)
Autofix CI daemonBriefbernstein autofix start/stop/status/attach - watches PRs, dispatches repair runs on CI failure (core/autofix/)
Dev previewBriefbernstein preview start/stop/list/status - exposes agent dev server via tunnel with configurable auth (core/preview/)
Fleet dashboardBriefbernstein fleet [--web HOST:PORT] - cross-session multi-instance view (core/fleet/)
Notification sinksBriefbernstein notify test --sink <id> - pluggable notification backends (core/notifications/)
PR review responderBriefbernstein review-responder start/status/tick - auto-responds to PR review comments (core/review_responder/)
Review pipeline DSLBriefbernstein review --pipeline review.yaml - YAML-driven multi-phase review (core/quality/review_pipeline/)
Plan archivalBriefbernstein plan ls/show - list and inspect archived plans (core/planning/lifecycle.py)
Slack integrationBriefSlash commands and events API endpoints
Webhook ingestionBriefPOST /webhooks/ for external event routing
Adaptive parallelismBriefcore/orchestration/adaptive_parallelism.py - auto-tunes concurrency from observed success rates
Warm poolBriefcore/agents/warm_pool.py - pre-spawned agent pool to cut spawn latency
Content-addressed artifact storeBriefcore/persistence/cas_store.py - content-addressed deduplication for artifacts
Workflow DSLBriefbernstein workflow validate/list/show
Chaos engineeringBriefbernstein chaos agent-kill/rate-limit/file-remove/status/slo
Benchmark suiteFullbernstein benchmark run/compare/swe-bench
Eval harnessBriefbernstein eval run/report/failures
SWE-Bench harnessFullVerified eval in benchmarks/swe_bench/run.py
Graduation systemBriefAgent promotion stages, routes in routes/graduation.py
Semantic cachingBriefsemantic_cache.py - prompt deduplication
Cascade router (intra-Claude tier escalation)BriefTier escalation within a single provider - see core/routing/cascade_router.py:386
Cascade fallback manager (cross-adapter failover)BriefCross-adapter provider failover - see core/routing/cascade.py:287
Batch routerBriefTask batching for non-urgent work
Prompt cachingBriefSHA-256 system prefix deduplication
Output style customizationBriefConfigurable agent output format
Installation mismatch detectionBriefDetects adapter/installation gaps
API preconnect warmupBriefConnection warmup before heavy runs
Worker badge identityBriefProcess identification in ps/Activity Monitor
Keybinding system (TUI)BriefConfigurable TUI keyboard shortcuts
Diff folding displayBriefFolded diff rendering in agent output
Word-level diff renderingBriefCharacter-level change highlighting
Contextual tips systemBriefIn-context hints for agents
Session tag systemBriefTag and filter runs
Rename sessionBriefSession renaming command
Security review commandBriefbernstein security-review
Commit attribution statsBriefPer-agent commit statistics
Away summary generationBriefSummarize what happened while you were away
Plugin trust warningBriefWarns on unverified plugins
Cumulative progress trackingBriefProgress tracking across runs

CLI commands

CommandDocs statusNotes
bernstein -g GOALFullInline goal
bernstein run plan.yamlFullPlan file execution
bernstein initFullWorkspace setup
bernstein stopFullGraceful/force stop
bernstein liveFullTUI dashboard
bernstein dashboardFullWeb dashboard
bernstein statusFullTask summary
bernstein psFullProcess list
bernstein costFullSpend breakdown
bernstein doctorFullPre-flight health check
bernstein recapFullPost-run summary
bernstein retroFullRetrospective report
bernstein trace IDFullDecision trace
bernstein logsFullAgent log tail
bernstein diff IDFullPer-task git diff
bernstein planFullTask backlog
bernstein replay IDBriefDeterministic replay
bernstein checkpointBriefSession snapshot
bernstein wrap-upBriefEnd session with summary
bernstein demoFullZero-config demo
bernstein quickstartBriefFlask TODO demo (3 tasks)
bernstein agents ...FullCatalog management
bernstein evolve ...FullSelf-improvement
bernstein ci fixFullCI autofix
bernstein github setupFullGitHub App setup
bernstein workerBriefJoin cluster as worker
bernstein mcpBriefRun as MCP server
bernstein chaosBriefFault injection
bernstein auditBriefCryptographic audit
bernstein verifyBriefMerkle/HAMC verification
bernstein benchmarkFullBenchmark suite
bernstein evalBriefEvaluation harness
bernstein ideateBriefCreative evolution
bernstein workspaceFullMulti-repo workspace
bernstein configBriefConfiguration management
bernstein quarantineBriefCross-run task quarantine
bernstein cacheBriefResponse cache management
bernstein test-adapterBriefAdapter smoke test
bernstein add-taskBriefInject task via CLI
bernstein cancelBriefCancel task
bernstein review/approve/reject/pendingBriefReview workflow
bernstein syncBriefSync backlog with server
bernstein manifestBriefRun manifest inspection
bernstein gatewayBriefMCP gateway proxy
bernstein workflowBriefWorkflow DSL
bernstein watchBriefDirectory file watcher
bernstein listenBriefVoice commands (experimental)
bernstein completionsBriefShell completion scripts
bernstein self-updateBriefUpgrade from PyPI
bernstein pluginsBriefList active plugins
bernstein install-hooksBriefInstall git hooks
bernstein debugBriefGenerate debug bundle for triage
bernstein acp serveFullACP bridge (--stdio or --http :PORT)
bernstein autofix ...BriefCI autofix daemon (start/stop/status/attach)
bernstein connectBriefCredential vault setup for a provider
bernstein creds ...BriefCredential management (list/revoke/test)
bernstein preview ...BriefDev server preview (start/stop/list/status)
bernstein fleetBriefFleet dashboard (optionally --web HOST:PORT)
bernstein mcp catalog ...BriefMCP catalog browser (browse/search/install)
bernstein notify testBriefNotification sink smoke test
bernstein plan ls/showBriefList and inspect archived plans
bernstein review-responder ...BriefPR review responder (start/status/tick)
bernstein review --pipelineBriefReview with YAML pipeline DSL

Cloud / Cloudflare

CapabilityDocs statusNotes
Workers RuntimeBridgeFullbridges/cloudflare.py - agents on Workers + Durable Objects
Workflow Bridge (durable execution)Fullbridges/cloudflare_workflow.py - auto-retry, approval gates
Sandbox Bridge (V8/container isolation)Fullbridges/cloudflare_sandbox.py - isolated code execution
Browser Rendering BridgeFullbridges/browser_rendering.py - screenshots, scraping, PDFs
R2 Workspace SyncFullbridges/r2_sync.py - content-addressed delta sync
Workers AI Provider (free LLMs)Fullcore/routing/cloudflare_ai.py - Llama, Mistral, Gemma, Qwen
D1 Analytics & BillingFullcore/cost/d1_analytics.py - usage metering, billing tiers
MCP Remote TransportFullmcp/remote_transport.py - streamable HTTP for remote MCP
Cloud CLI (bernstein cloud)Fullcli/commands/cloud_cmd.py - login, run, status, cost, deploy
Cloudflare Agents AdapterFulladapters/cloudflare_agents.py - wrangler dev integration
Codex-on-Cloudflare AdapterFulladapters/codex_cloudflare.py - Codex in CF sandboxes