evo-agents
English | Tiếng Việt
A shared knowledge base and plugin workspace for AI agent teams.
evo-agents brings your codebases, documents, chats, specifications and business signals into one permission-aware hub, extends that workspace across product roles with plugins, and dispatches verified work to Claude Code, Codex CLI or opencode on machines you control.
Get started | Website | Plugin guide | Documentation
Project knowledge base on the hub
Every project keeps one shared knowledge base on the hub that serves both people in the browser and agents over MCP.
- Bring code and non-code sources together. Index Git repositories (code symbols across about 40 languages, commit history and pull requests), drop
.zipor.tar.gzarchives of documents and spreadsheets in the browser (New project from dataorUpload data), include approved SRS use cases, Lean Canvas claims and group chat threads, or connect external wikis and tools. Secrets and API keys are redacted before indexing. - Link concepts, specs and code with role-aware visibility. The graph connects domain concepts and document passages directly to files, functions, plan steps and commits. Every member and agent sees only what their clearance and chat membership allow. Named graphs let you scope each agent to a chosen subset of sources, while scheduled optimization runs evaluate retrieval quality and roll back regressions automatically.
- Keep versioned docs, memories and direct answers. Teams write versioned Markdown documents and sandboxed interactive pages in Docs, share skills and memories across sessions, and ask questions directly in Home chat.
Workspace extended by plugins
A product team needs more than a queue for code commits. Plugins turn the hub into one shared workspace for business analysts, product managers, strategists, marketers, editors and engineers by slotting pages, specialist agents, graph sources and workflow rules into the project's Business cycle and Development cycle.
- Extend both product cycles. Built-in and core plugins add living Lean Canvases (where claims reach validated only with real evidence and signals), versioned SRS specifications reviewed by
@ba, linked Tasks and daily PM digests, inbound CRM contacts, gated Marketing channels, OKR trees and Editorial pipelines. - Core and sandboxed plugins on one model. Plugins ship in two tiers: core modules bundled with the hub, and third-party isolated plugins that run in separate OS processes and sandboxed UI frames, acting only within the permissions each project grants (plugin guide).
- Install from trusted markets. Hubs read decentralized plugin markets on GitHub, verify publisher signatures before installing free releases, and support per-hub licences for commercial plugins (market guide). Teams can scaffold custom domain plugins from the template or with
@builder.
Agent systems and verified execution on your machines
Underneath the knowledge base and domain plugins, multi-agent systems and worker daemons turn plans and requests into verified changes on hardware you own.
-
Wire multi-agent systems visually. Connect an entry point, specialist agents, an Orchestrator that manages queues and stalled plans, a nightly Curator that proposes and tests workspace improvements, and named knowledge graphs. Each agent runs on its own dedicated worker and quota pool.
-
Hand off whole plans with re-verified checks. Describe a change on the Plans page, for example:
Add saved carts to checkout. Cover the API and frontend, and include tests for restoring a cart after sign-in.
An agent drafts the steps and checks; once started, a worker runs each step in an isolated Git worktree using Claude Code, Codex CLI or opencode, re-runs its verification commands before marking it done, and pushes commits on the plan's branch.
-
Keep credentials sealed and approvals in your hands. Store model subscriptions, API keys and Git tokens on the hub as sealed secrets leased per run. Your autonomy settings pause a run only on decisions that require human approval in the Inbox or on Telegram.
Quick start
Join or host a hub. Request an invite, sign in to your team's hub, or host one on a single machine. Install Python 3.11+ and uv, then sign in:
uv tool install 'evo-ak[graphify,worker]'
evo-agents hub login --url https://hub.example.org
Replace https://hub.example.org with your hub's URL. From there you can:
- Start from documents or spreadsheets in the browser with New project from data (
/projects/new), dropping a.zipor.tar.gzarchive to build a searchable knowledge base without setting up a repository first (guide). - Register repositories and connect a worker with
evo-agents harness init,evo-agents hub project registerandevo-agents worker joinfollowing the getting started guide. Workers run on macOS, Linux, or Windows under WSL.
Connect your terminal coding agent. Install the client plugins so interactive sessions read and write hub plans, memories and the knowledge graph over MCP, spool oversized tool outputs locally, and attach call-graph hints after grep:
# Claude Code
claude plugin marketplace add https://github.com/maycuatroi1/evo-agents
claude plugin install evo-hub@evo-agents
claude plugin install evo-kg@evo-agents
# Codex CLI
codex plugin marketplace add https://github.com/maycuatroi1/evo-agents
codex plugin add evo-hub@evo-agents
The PyPI distribution is evo-ak and the CLI command is evo-agents. evo-agents is pre-1.0: commands, APIs and schemas can change between releases. Check the roadmap for current status.
Documentation
| Topic | Guide |
|---|---|
| Set up a project, upload data and run a plan | Getting started |
| Configure knowledge sources and query tools | Graph setup, query reference, evaluation |
| Build, isolate, sign and publish plugins | Plugin protocol and host, Plugin market |
| Product cycle workflows (Canvas, SRS, Tasks, Loop) | Problems, Canvas, SRS, Tasks, Loop |
| Run workers, manage credentials and set autonomy | Operations, workers, credentials, autonomy |
| Self-host or deploy a branded client hub | Single server, hub deployment |
| Contribute or report a security issue | Contributing, Security policy |
Agent instructions and MCP reference live in AGENTS.md and the agent guide.
License
Metadata
Release files for evo-ak 0.70.5
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 | Actions |
|---|---|---|---|---|
| evo_ak-0.70.5.tar.gz | 5.1 MB | With provenance | Details for evo_ak-0.70.5.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| evo_ak-0.70.5-py3-none-any.whl | Python 3 | none | any | Details for evo_ak-0.70.5-py3-none-any.whl |
Total release size: 11.3 MB
Release files / evo_ak-0.70.5.tar.gz
| Download URL | evo_ak-0.70.5.tar.gz |
|---|---|
| Size | 5.1 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
db182903332c5144fcb717e406fec1736cd7a77b274c607e8a4c6572b83df777
|
|
BLAKE2b-256 checksum How to use checksums |
fefbd11bd9f3363317f615e37ad1acaa44c9e8668570fadf36f9fc1d7b03b39f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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 Nov 18, 2026.
Transparency logRelease files / evo_ak-0.70.5-py3-none-any.whl
| Download URL | evo_ak-0.70.5-py3-none-any.whl |
|---|---|
| Size | 6.1 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
ff0d4d7a72ee592c755a34a44b614a5785bde62eb580f303a6e364c52021d8a9
|
|
BLAKE2b-256 checksum How to use checksums |
1d0aa960ec63f3bbe4a0b484fbebf4e12554893c9c4a53db280d305249253623
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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 Nov 18, 2026.
Transparency logSecurity for evo-ak 0.70.5
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
Both files in this release have been uploaded with provenance metadata.
Attestation for evo_ak-0.70.5.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 Nov 18, 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.
To learn more about protecting yourself from malicious actors, visit our package security documentation.