greatminds
Greatminds coordinates coding agents through ACP. A local daemon owns task assignment, sessions, process cleanup, validation commands, and workflow transitions. Agents perform work that needs reasoning and submit typed results. Task files and evidence remain on the filesystem; there is no database service.
The local web workspace provides role tabs,
dashboards, settings, and batch-run inspection without application authentication.
Start it with greatminds web --project-dir /path/to/project --port 8765.
Quickstart
The managed daemon runs on Linux with Python 3.11+, /proc and pidfd support.
See the host requirements.
pip install greatminds # or install in your preferred Python environment
mkdir -p /tmp/greatminds-demo
cd /tmp/greatminds-demo
greatminds setup
Setup creates coordination/execution.yaml with empty agents and bindings,
and the queue tree under .greatminds/. It does not install harnesses, change
authentication, install plugins or services, or choose a provider for you.
Repeated setup preserves your execution contract and task data.
Local ACP workflows do not require Ansible or a deployed stand. If you enable YAML stand profiles, install the optional dependency in the same environment:
python -m pip install 'greatminds[stands]'
This adds the validated Ansible version range; it does not configure hosts or authorize deployment. Required stand/evidence gates still apply to tasks that declare them.
Add an ACP agent and a role binding to coordination/execution.yaml. This is a
configuration example: replace the executable path and version labels with your
installed adapter and harness, and complete that harness's authentication first.
version: 1
agents:
local-agent:
transport: acp
argv: [/absolute/path/to/your-acp-agent]
adapter_version: your-tested-adapter-version
harness_version: your-tested-harness-version
bindings:
planner:
agent: local-agent
role: ARCHITECT-PLANNER
scheduling: on-demand
permission: ask
Role, harness, model, permissions, workspace and scheduling are independent.
Use scheduling: queue for automatic task assignment. Additional bindings can
use the same agent manifest or a different ACP implementation. See the
execution contract for command,
worktree, model, account and stand policies.
An existing explicit contract can be supplied during initialization:
greatminds setup --project-dir /path/to/project --execution-config /path/to/execution.yaml
Start the independent background daemon (the command returns when ready):
greatminds coordd
On Linux with a systemd user manager, service installation is optional:
greatminds daemon install --name my-project
greatminds daemon start --systemd --project my-project
The registry associates the service name with the project directory. Without an
explicit name, installation uses an existing registered name or the directory
name. Names accept ASCII letters, digits, dots, underscores and hyphens (1–80
characters, starting with a letter or digit). Use an explicit name for directories
containing spaces or when another project has the same basename. Existing names
cannot be redirected to a different project. No coord.yaml is required.
The service uses a generic PATH containing the launcher directory, standard
system directories and user bin directories. For additional runtime locations,
set PATH in .greatminds/PROJECT.env, or declare an agent/command environment
reference for PATH in execution.yaml. Service installation does not run login
shells or discover vendor-specific executables.
greatminds update refreshes the ACP project and only already-installed services.
It uses try-restart, so inactive services stay inactive. It never installs a
service or opens a frontend. Restart a foreground daemon manually after updating.
update --project NAME selects the registered project even from another directory.
In another terminal, create an operator conversation with the configured binding:
greatminds chat create planner
# Use the conversation ID returned above:
greatminds chat talk CONVERSATION_ID
The daemon owns the ACP session. Detaching leaves the work running; reconnecting
does not replay completed prompts. talk displays streamed text and explicit
one-time permission choices. chat close CONVERSATION_ID requests cleanup.
Task-bound conversations use chat create BINDING --task TASK_ID.
greatminds run status
greatminds run events --follow
greatminds run pause
greatminds run resume
greatminds run cancel RUN_ID
greatminds run retry RUN_ID
Local frontends
greatminds launch --target tmux
greatminds launch --target vscode
Tmux hosts the daemon and an operator shell. The VS Code target writes
greatminds-acp.code-workspace with process tasks for the daemon and operator
commands, preserving unrelated workspace settings. Neither frontend launches
native harness TUIs. The VS Code extension supplies
New ACP Chat, Attach ACP Chat and Close ACP Chat commands.
Harness compatibility
Claude Code and OpenAI Codex use upstream ACP adapters. Other harnesses may expose ACP directly or through a bridge. Protocol initialization alone does not establish working authentication, session continuity, model selection or tool permissions. The compatibility matrix distinguishes verified scenarios from untested support for Codex, Claude, Grok, Qwen, Kimi, Cline, Gemini, OpenHands and Cursor. The same client handles supported harnesses; there is no fallback to a native transport.
Tasks, evidence and stands
Task queues live under .greatminds/; editable project configuration lives under
coordination/. The installed schema is authoritative and is pinned to each run.
Successful ACP output alone never completes a workflow transition: the daemon
validates typed results, task revision, permissions and evidence first.
Configured validation commands run through the daemon with bounded output and
recorded receipts. Optional stand deployment uses an explicit policy in the
execution contract and profiles in coordination/stand-profiles.yaml and
coordination/stand-profiles/. Ansible-backed profiles remain available; setup
does not enable deployment or seed a deployment policy automatically.
greatminds stand profiles list
greatminds stand profiles doctor
Development status
The primary setup, launch and daemon entrypoints use ACP. The modernization acceptance covers Codex, Claude and Grok; see the acceptance record for scope and reproducible evidence. The local web workspace is included in the 3.0 release. See the upgrade guide when moving from 2.x.
Issues and license
Report Greatminds defects in the project issue tracker. Bugs in a project managed by Greatminds belong in that project's tracker.
Apache-2.0. See LICENSE.
1 maintainer
Metadata
1 maintainer
Release files for greatminds 3.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| greatminds-3.1.0.tar.gz | 288.8 kB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| greatminds-3.1.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 631.8 kB
Release files / greatminds-3.1.0.tar.gz
| Download URL | greatminds-3.1.0.tar.gz |
|---|---|
| Size | 288.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
c5553e14ba57dda76c8e0c83889f675780db4f61ef3e08e39a0da80ed83d473a
|
|
BLAKE2b-256 checksum How to use checksums |
c7e6ad9f8fa46fe514eda91a6591c11a0d5d6f17b9af501a300e3e9f3b2d350d
|
| 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 Sep 7, 2026.
Transparency logRelease files / greatminds-3.1.0-py3-none-any.whl
| Download URL | greatminds-3.1.0-py3-none-any.whl |
|---|---|
| Size | 343.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
dfa106724a5db2b25f651cd6afd9de401b867cb6a628b90ee635a7fd63272e9b
|
|
BLAKE2b-256 checksum How to use checksums |
0610e05c8ff8b0b1820d024c88c3589db333c68ef9e84872dbacfca04a68cb16
|
| 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 Sep 7, 2026.
Transparency logSecurity for greatminds 3.1.0
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.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for greatminds-3.1.0.tar.gz
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 Sep 7, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
