fkteams

Coordinate specialist agents to handle demanding development, research, and operations work.

Stars
★ 153
Last updated
1mo ago
License
MIT
Primary language
Go

At a glance

How it runs
CLIWeb appSelf-hosted service
Works with
Universal · cross-platformOpenAI API
Cost
Free software; you pay for model usage
Setup effort
Medium · a few setup steps
You'll need
GoBunShell / CLINetwork accessLocal filesystemMCP Server
Typical use
A developer wants specialist agents to split a complex coding task and needs to track their progress in a Web UI.
Not a fit if
  • Users who only need a single chat assistant
  • Teams expecting model access without configuration

What does this agent do, and when should you use it?

fkteams is a locally runnable multi-agent assistant whose coordinator assigns work to specialists in coding, research, analysis, and remote operations. Start tasks through its Web UI, CLI/TUI, or API service, or connect Discord, QQ, and WeChat channels. It offers team, deep-work, and roundtable discussion modes, with background execution, reconnect recovery, live redirection, and queued follow-up questions. Built-in tools cover files, commands, search, documents, spreadsheets, Git, and SSH; MCP, Skills, JavaScript, custom agents, and workspace rules provide extension points. Sessions, configuration, and workspaces are stored locally by default, while model access is configured through a login wizard for common OpenAI-compatible services and GitHub Copilot.

Install fkteams and run fkteams login to configure a model. Start the Web UI with fkteams web, use the CLI/TUI by running fkteams, or launch an API service with fkteams serve. After a user submits a coding, research, analysis, or operations task, the coordinator assigns it to specialist agents. Those agents can read and write workspace files, run commands, search for information, and use document, spreadsheet, Git, and SSH tools. The interface exposes task progress and tool calls, and long-running work can continue in the background, resume after disconnects, and accept follow-up questions. Add capabilities through Skills, MCP, custom tools and workflow Hooks, custom agents, and workspace AGENTS.md rules.

  1. A developer wants specialist agents to split a complex coding task and needs to track their progress in a Web UI.
  2. A researcher needs agents to search for information, organize documents, and continue work over a long session.
  3. An analyst wants one task workflow to read files, work with spreadsheets, and call analysis tools.
  4. An operations engineer needs command and SSH tools for remote work, with visibility into tool calls.
  5. A team wants to connect a locally run task assistant to an API or Discord, QQ, or WeChat channel.

How do you install or deploy this agent?

Install on Linux or macOS with the shell script, or on Windows with PowerShell. Manual downloads are also available from GitHub Releases. Building from source requires Go and Bun:

curl -fsSL https://raw.githubusercontent.com/wsshow/feikong-teams/main/install.sh | bash
powershell -c "irm https://raw.githubusercontent.com/wsshow/feikong-teams/main/install.ps1 | iex"

Build from source:

git clone https://github.com/wsshow/feikong-teams.git
cd feikong-teams
make native

How do you use this agent?

Configure a model, then start the Web UI:

fkteams login
fkteams web

Open http://localhost:23456 to create a multi-agent task. For terminal work, run fkteams to start the CLI/TUI. For integrations and automation, run fkteams serve to start the API service. The model login wizard supports common OpenAI-compatible providers and GitHub Copilot.

What are this agent's strengths and limitations?

Pros
  • A coordinator assigns work to coding, research, analysis, and remote-operations specialists, with team, deep-work, and roundtable modes.
  • The Web UI, CLI/TUI, and OpenAI-compatible API share session and execution capabilities; Discord, QQ, and WeChat channels are also supported.
  • Background execution, reconnect recovery, live redirection, and queued follow-ups support long-running tasks.
  • Built-in file, command, search, document, spreadsheet, Git, and SSH tools can be extended through MCP, Skills, and JavaScript.
  • Sessions, configuration, and workspaces stay local by default; high-risk tools can use permission policies, human confirmation, and execution audits.
Limitations
  • Users must configure a model with fkteams login; the README does not state model costs or promise a free model.
  • Building from source requires Go and Bun, and model setup is needed before productive use.
  • The available tools include command execution, network search, and SSH, so adopters need to review tool permissions and data access.
  • The README provides no measured success rates, resource requirements, or comparison of features across model providers.

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
fkteams This agent 40 · Major gaps CLIFree + model costs ★ 153 1mo ago Go OpenAI API
CCCC Coordination Console 61 · Some gaps CLIFree + model costs ★ 1.3k today Rust Codex · Claude Code
CC-Connect 60 · Some gaps CLIFree + model costs ★ 16k 11d ago Go Codex · Claude Code
Agent Toolkit 74 · Some gaps CLIFree ★ 19 1d ago V ChatGPT · Codex · Claude Code

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
40/ 100 5-point scale 2.0 / 5
Trust 7/29
Reliability 6/14
Adaptability 9/18
Convention 8/18
Effectiveness 7/13
Verifiability 3/8
Why each dimension lost points
Trust7 / 29 · 1.2/5

The README says sessions, configuration, and workspaces are local by default, and mentions permission policies, human confirmation, and audit records for high-risk tools. These provide some support for controls. The same document lists broad capabilities including commands, network access, SSH, MCP, scripts, and messaging, while the supplied material does not detail permissions, data transfer paths, sensitive-data protections, or rollback and recovery. The dependency list has no security-audit evidence. The acknowledgements name several upstream projects but do not establish maintainer identity or a complete provenance chain, so these criteria receive only partial credit.

Reliability6 / 14 · 2.1/5

The README's descriptions of interfaces, collaboration modes, and long-running tasks are broadly consistent. The CI configuration shows that pushes and pull requests run make check, which provides some support for self-consistency. The material does not show what that check verifies or describe failure handling. Ordinary use also depends on Go, Bun, model services, and external integrations, so dependency availability and failure messages receive low scores.

Adaptability9 / 18 · 2.5/5

The README names software development, research, analysis, remote operations, and automation as scenarios, describes team, deep-work, and roundtable modes, and lists Web, CLI/TUI, API, and messaging interfaces. It also mentions installation on Linux, macOS, and Windows and Docker deployment. The supplied material does not specify capability boundaries, mode-selection guidance, or trigger rules, so those criteria receive only partial credit.

Convention8 / 18 · 2.2/5

The README organizes content into quick start, usage, extensions, and documentation topics, with installation commands, configuration entry points, and documentation links. The supplied material contains only basic command examples and no FAQ, end-to-end example, or known-limitations section. The complete MIT license earns full credit. A release badge and a tag-triggered release workflow indicate a versioning path, but no changelog is provided. Publisher identity is unknown, and responsibility for maintenance and updates is only lightly described.

Effectiveness7 / 13 · 2.7/5

The product description presents collaboration modes, multiple interfaces, background task tracking, follow-up queues, and extensible tools, showing plausible value for complex long-running tasks and giving users clear entry points. The supplied material includes no output examples or cost guidance. Multi-agent work, model services, and numerous integrations may add usage and operational costs, so the benefit and cost criteria receive partial scores.

Verifiability3 / 8 · 1.9/5

The README links feature claims to configuration, deployment, security, architecture, and API documentation, and lists dependencies and a CI workflow, providing some traceability and corroboration. The linked document contents are not supplied, and the CI file only shows that make check is invoked; it does not establish test coverage or results. Product capabilities are mostly project claims, with limited separation between facts and inference, so these criteria receive partial scores.

Risks and how to mitigate them
  • Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
  • Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
  • This assessment uses only the supplied README, license, Go dependency manifest, and GitHub Actions configuration; implementation code was not reviewed, and neither the project nor CI was run.
  • The project offers command, network, SSH, scripting, and external-service integrations. Before deployment, consult the referenced security and configuration documentation to establish permission scope, credential protection, data flows, and recovery options.
Evidence confidence: Low Reviewed Oct 10, 2026 Reviewed revision 8b9fcc38e277
See the full review method →

FAQ

Do I need a paid model service?
The software is under the MIT License, and the login wizard supports common OpenAI-compatible services and GitHub Copilot. The README does not state model pricing or promise a free model, so check the selected provider's costs.
Where is task data stored?
The README says sessions, configuration, and workspaces are stored locally by default. See the configuration guidance for deployment and authentication settings.
Can it run commands or access remote machines?
Its built-in tools include command execution and SSH. The README says high-risk tools can use permission policies, human confirmation, and execution audits.
Can I use it from a terminal or another program?
Yes. Run fkteams for the CLI/TUI or fkteams serve for the API service; the README also identifies an OpenAI-compatible API entry point.
Will a long-running task be lost if I disconnect?
The README documents background execution and reconnect recovery. For Web authentication, logging in again after expiration restores the task event stream.
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related agents