Skip to main content

ChaseOS Core

tests License: MIT Python 3.11+ Status: Alpha Linted with Ruff

ChaseOS Core is an MIT-licensed, local-first framework for building governed hybrid-intelligence operating systems. It gives fork owners a safe scaffold for memory, projects, source intake, runtime boundaries, approval-gated automation, agent coordination, and evidence-first writeback across humans, deterministic software, ML, and generative agents.

Core is the public framework layer. Your private identity, projects, logs, credentials, provider state, and live runtime memory belong in a separate private ChaseOS instance.

Why ChaseOS Core

Most agent frameworks optimize for capability. Core optimizes for authority — what an automated system is allowed to do, who is accountable for it, and what evidence exists afterward. That shows up as three concrete defaults:

  • Fail-closed, not fail-open. Adapters without a bound backend deny rather than proceed. Provider connections start read_only; writes and external egress are approval-gated.
  • Modality before provider. The Autonomous Operator Runtime decides whether a step belongs to a human, deterministic code, an ML model, or a generative agent before it picks a runtime — and can derive the approval plan without executing anything.
  • Truth is promoted, not written. Captures land in quarantine; canonical knowledge is reached through an explicit review gate, not direct agent writeback.

Authority pipeline: a request is routed by modality, gated by the ApprovalGateway port, executed within a bounded write scope, and evidenced. Without a bound gateway the path denies by default, and canonical knowledge is reached only through an explicit promotion gate.

If you're evaluating the design, start with docs/ARCHITECTURE.md.

Documentation

Doc What it covers
Architecture Layer model, module map, decision-routing and connections flows (diagrams)
Decision records (ADRs) Why the architecture is shaped the way it is
Quickstart Fork-first setup path
Core Operating Model The seven canonical layers
Command Reference Full CLI surface
Permission Matrix Trust tiers and authority ceilings
Approval Center Approval-gated write governance
FORKING.md What to keep standard vs. customize in a fork
CONTRIBUTING.md Dev setup, scope rules, PR expectations
Releasing PyPI setup and how a release is cut
examples/ Runnable scripts: use Core as a library in your own project
CHANGELOG.md Release history

What Works Today

Area Current Core capability Boundary
Repository framework Public folder structure, governance docs, templates, SOPs, runtime standards, and example files Examples are scaffolds; private content is not included
Lean Core CLI chaseos version, chaseos doctor, chaseos capture, chaseos schedule, chaseos run, chaseos connections, chaseos commerce The Core CLI is intentionally smaller than any private/proprietary operator CLI
Health checks chaseos doctor --vault-root . --json validates importable Core runtime blocks and the vault root shape Checks for vault markers (00_HOME/, .chaseos/) and reports which matched; it does not validate private provider accounts or credentials
Capture intake Explicit file, stdin, and optional local image-text capture into quarantine/intake paths No ambient screen capture, browser-profile capture, cloud OCR, or canonical promotion by default
Connections registry Local-first provider manifest discovery plus SQLite registry initialization/seeding Does not authenticate providers, fetch private data, or send messages by default
Bounded workflow runner chaseos run <workflow_id> --dry-run routes through the AOR/workflow substrate Core ships no workflow manifests, so this resolves to escalated rather than executing. Running real workflows requires supplying your own registry — see ROADMAP.md
Decision-route inspection chaseos decision-route inspect <contract.json> --json deterministically checks human/rules/ML/genAI step selection and derives a decision-scoped approval plan Inspection only: no dispatch, approval consumption, permission change, credential access, or canonical writeback
Schedules Native schedule-intent listing Core does not ship private schedules or live operator queues
Read-only commercial foundation Catalog, entitlement, flags, admin overview, and ledger surfaces Intended as product/marketplace scaffolding; not production billing by itself
Runtime governance Permission matrix, trust tiers, adapter standards, Gate interface, task routing, role-card patterns Governance docs and code are framework-level; private deployment policy belongs outside public Core
Repo-safe secret audit runtime.repo_secret_audit can scan tracked/untracked repo text without emitting raw secret values It is a safety check, not a substitute for manual review before publishing

AOR and Approval Preflight

The Autonomous Operator Runtime is not agent-only orchestration. Its expanded contract selects the appropriate modality for each material step before runtime/provider selection:

  • human for accountability, ambiguity, liability, and protected decisions;
  • rules/code for exact, stable, security-sensitive, or authority-bearing operations;
  • ML/statistics for versioned prediction over structured historical data;
  • generative AI for bounded interpretation, synthesis, language, and flexible planning.

The first Core foothold is a fail-closed, read-only Decision Modality Router:

chaseos decision-route inspect docs/runtime/Decision-Contract.example.json --json

It validates the route and produces an approval plan naming the accountable human, exact decision scope, reasons, required evidence, and block-on-timeout/denial behavior. It does not execute the route. See 06_AGENTS/Autonomous-Operator-Runtime.md and docs/governance/Approval-Center.md.

Scaffold / Experimental Surfaces

These surfaces are present as Core-safe scaffolds or bounded footholds. Treat them as building blocks, not fully enabled production integrations:

  • provider manifests for Discord, Telegram, Slack, WhatsApp Business Cloud, WhatsApp personal lab, iMessage Mac, GitHub, and local files;
  • Studio product-surface contracts and launch references for apps built above Core;
  • adapter manifests and runtime profile examples for external agent runtimes;
  • acquisition/source-pack, graph, memory, subagent, OSRIL, operator-surface, and AOR modules;
  • optional browser and voice extras declared in pyproject.toml.

What This Repository Contains

  • Framework documentation for the ChaseOS control plane.
  • Templates for notes, projects, logs, runtime profiles, approvals, and audits.
  • Governance patterns for approval-gated writes and evidence-first promotion.
  • Adapter standards for external runtimes and model/tool surfaces.
  • Example folders that can be copied into a private deployment.
  • The lean MIT Core CLI entrypoint: chaseos = runtime.cli.core_main:main.
  • Local-first runtime modules for capture, schedules, connections, commerce scaffolding, AOR, Gate interfaces, graph/memory inspection, and bounded operator surfaces.
  • Studio product-surface contracts for the application layer built above Core.

What This Repository Intentionally Does Not Contain

  • Personal notes or private project state.
  • Live runtime logs, approval queues, agent-bus state, or private schedules.
  • Credential values, API keys, tokens, private keys, or provider secrets.
  • Provider-specific deployment state or authenticated account bindings.
  • Machine-local paths or operator-specific usernames.
  • Proprietary/private Studio builds or packaged installers.
  • Public write authority to canonical knowledge without review/promotion.

Intended Use

Use Core as a starter kit and reference model. Private deployments should keep local content, runtime state, and operator records outside the public Core tree.

A healthy fork should:

  1. keep chaseos-core as framework/source;
  2. create a separate private ChaseOS instance for personal or business operations;
  3. copy templates and examples into that private instance only when needed;
  4. store credentials in ignored local stores or external secret managers;
  5. run read-only/dry-run checks before enabling write, connector, browser, or publication actions;
  6. promote durable truth through an explicit review gate instead of direct agent writeback.

ChaseOS Studio should be treated as an application layer over these Core contracts. Ship public contracts and reviewed source-safe docs in the repo; distribute packaged installers such as .exe files through release channels rather than normal source commits.

Use Core in your own project

You do not have to fork Core to use its governance primitives. Install it as a dependency and import what you need — the modality router, the Gate port, and manifest loading all work as plain library calls with no vault directory:

pip install chaseos-core
from runtime.gate_interface import check_runtime_operation, register_gate

# With no provider registered, Core denies by default — it never silently permits.
allowed, reason = check_runtime_operation("vault.write")
assert allowed is False

class MyPolicy:
    def check_runtime_operation(self, operation, **kwargs):
        return operation.endswith(".read"), f"policy decision for {operation}"
    # ... plus the remaining GateProvider methods

register_gate(MyPolicy())

Runnable scripts for the three main entry points are in examples/, and examples/README.md documents which parts of Core need a vault root and which do not.

Quickstart

See docs/getting-started/Quickstart.md for the fork-first setup path.

Minimal validation commands:

python -m venv .venv
source .venv/bin/activate  # Windows PowerShell: .venv\Scripts\Activate.ps1
pip install -e .
chaseos version
chaseos doctor --vault-root . --json
chaseos connections providers --json

Run a repo-safe secret audit before publishing a fork:

python - <<'PY'
from runtime.repo_secret_audit import audit_repo_secrets, format_repo_secret_audit
print(format_repo_secret_audit(audit_repo_secrets('.')))
PY

Contributing

Contributions are welcome. See CONTRIBUTING.md for dev setup, scope rules (what belongs in Core versus a private instance), and PR expectations. This project follows the Contributor Covenant.

Security issues should not be filed as public issues — see SECURITY.md.

License

ChaseOS Core is released under the MIT License. See LICENSE.md. Third-party dependency notices are recorded in THIRD_PARTY_NOTICES.md.

Release files for chaseos-core 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for chaseos-core 0.1.0
File Size Uploaded
chaseos_core-0.1.0.tar.gz 389.6 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for chaseos-core 0.1.0
File Interpreter ABI Platform
chaseos_core-0.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 872.6 kB

Release files / chaseos_core-0.1.0.tar.gz

Download URL chaseos_core-0.1.0.tar.gz
Size 389.6 kB
Tags Source
SHA-256 checksum
How to use checksums
c44ad799604bc1bd6deaf956d87ecda81db661fe925ead681388d600dccead14
BLAKE2b-256 checksum
How to use checksums
d9d522bfd092ab00d6af362466118900edb98ac9c270585b0fd9ac1db0632a35
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Aug 5, 2026.

Transparency log

Release files / chaseos_core-0.1.0-py3-none-any.whl

Download URL chaseos_core-0.1.0-py3-none-any.whl
Size 483.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
ce95d5cddca040e0b3dbcee509806a7610a5c4bfba8b25987e63210c9c8e94ad
BLAKE2b-256 checksum
How to use checksums
c177ef59608e094b17863d95bf0df2d704879a0c644ef790c2242db39faac4c1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Aug 5, 2026.

Transparency log

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page