LOCAL DAEMON API
Run a persistent service managing AI orchestration with full control over your infrastructure.
// SYSTEM STATUS: OPERATIONAL DAEMON: RUNNING PREVIEW: ACTIVE
Security-first
orchestration
for serious
local AI work.
CerbiBot is a security-first local AI orchestration platform with a daemon, operator dashboard, governed skills, project-scoped workflow control, and visible execution state. More control than a chat app. Less chaos than a loose agent stack.
[CERBIBOT-DAEMON]
RUN workflow-2847 [ACTIVE] ● ──────────────────────────────────────────────────────────────────────────── mode consensus agents 3 / 3 responding heartbeat 1.8s ago [healthy] ● provider anthropic / claude-3-7 [routed] ● skill research-deep [approved] ● memory scope project-alpha [isolated] ● CHECKPOINT ───────────────────────────────────────────────────────────────── consensus_reached 3 agents agreed on synthesis approach ARTIFACT ─────────────────────────────────────────────────────────────────── analysis-report.md [captured] ● RUN workflow-2848 [ACTIVE] ● ──────────────────────────────────────────────────────────────────────────── mode critique agents 2 / 2 responding heartbeat 0.9s ago [healthy] ● provider openai / gpt-5.4 [routed] ● skill patch-review [approved] ● memory scope infra-audit [isolated] ● CHECKPOINT ───────────────────────────────────────────────────────────────── critique_resolved blocking issue cleared before artifact release ARTIFACT ─────────────────────────────────────────────────────────────────── deployment-plan.md [captured] ●
// WHAT IT IS
Built for control.
CerbiBot combines a local daemon API, an operator dashboard, CLI control, and governed execution for skills and risky actions. It is designed for people who want more control than a chat app and less chaos than an unstructured agent stack.
OPERATOR DASHBOARD
Visual interface for monitoring runs, reviewing checkpoints, and managing provider configurations.
CLI CONTROL
Full command-line interface for scripting, automation, and integration with existing toolchains.
GOVERNED EXECUTION
Skill manifests, signing, validation, and approval-aware execution for stronger control.
// SECURITY AND CONTROL
Operator control by design.
[SECURITY-MODEL]
API ACCESS ............. bearer-token protected [●] OK SENSITIVE ACTIONS ...... admin-password-gated [●] OK PROJECT ISOLATION ...... per-project memory and session scope [●] OK SKILL VALIDATION ....... manifest-required, approval-aware [●] OK AUDIT VISIBILITY ....... full run log, checkpoints, blockers [●] OK EXPORT CONTROLS ........ export-aware execution paths [●] OK
Structured Orchestration
Multiple execution modes handle different AI collaboration patterns with explicit operator control.
Visible Execution
Runs, checkpoints, heartbeats, blockers, and recovery state remain visible instead of opaque.
Reusable Skills & Artifacts
Skills and captured artifacts persist across workflows while staying bounded by operator-approved rules.
// EXECUTION MODES
Multiple modes. One system.
[MODE: SINGLE]
A single operator agent handles the task directly with the full local toolchain and audit trail intact.
- agents
- 1
- output
- direct execution artifact
- checkpoint
- single-run completion gate
- use cases
- fast fixes, scripting, deterministic tasks
[MODE: CRITIQUE]
One agent produces a draft and a second pass challenges assumptions before the final artifact is released.
- agents
- 2 staged passes
- output
- draft plus critique record
- checkpoint
- critique must resolve blockers
- use cases
- code review, risky edits, compliance checks
[MODE: RETRIEVAL]
Execution is grounded in project or external sources before synthesis, keeping responses tied to evidence.
- agents
- 1 retrieval-driven path
- output
- source-backed response
- checkpoint
- retrieval coverage validated
- use cases
- docs lookup, research, repo audits
[MODE: DEBATE]
Competing approaches are argued in parallel so tradeoffs are visible before a decision is selected.
- agents
- 2-4 opposing roles
- output
- tradeoff summary
- checkpoint
- winner selected after challenge
- use cases
- architecture choices, planning, validation
[MODE: CONSENSUS]
Multiple agents process the task independently, then converge on a synthesized response through a structured agreement step. Appropriate for high-stakes decisions where single-agent output carries risk.
- agents
- configurable (2-6 recommended)
- output
- synthesized consensus artifact
- checkpoint
- agreement threshold enforced before completion
- use cases
- analysis, research, validation, audit
[MODE: COUNCIL]
Specialized agents with distinct responsibilities report into a coordinating operator for controlled multi-role execution.
- agents
- role-based council
- output
- coordinated multi-agent plan
- checkpoint
- role approvals captured
- use cases
- large projects, orchestrated delegations, oversight
// TARGET OPERATOR
Built for people who run things.
- Technical operators running local AI workflows
- Builders who want repeatable, auditable orchestration
- Developers who need execution visibility, not just outputs
- Teams where governance and project isolation matter
- Anyone who’s outgrown unstructured agent chaos
// CURRENT STATE
Public technical preview.
STATUS: OPERATIONAL
- single, critique, retrieval, debate, consensus, council modes
- Skill validation, testing, and signing
- Run tracking with checkpoints and heartbeats
- Project isolation and scoped memory
- Artifacts, audit logging, and export controls
- Dashboard + CLI
- Provider controls and model routing
- Delegation and remote-access surfaces
// VALIDATION — CURRENT PREVIEW
- Live acceptance testPASS
- Local pentestPASS
- Dashboard typecheckPASS
- Production buildPASS
// INSPECT THE STACK
One repo. Full visibility.
CerbiBot is published as a single monorepo so you can inspect the full local stack in one place: backend, dashboard, docs, and scripts.
[SETUP]
$ git clone [email protected]:RavenFrostStudios/Cerbibot.git $ cd Cerbibot $ bash scripts/run_local_stack.sh // monorepo layout backend/ orchestration daemon and APIs dashboard/ cerbibot operator UI docs/ feature and launch documentation scripts/ local startup and shutdown