Productivity & Collaboration model-managementskill-managementmcp-managementprompt-managementdesktop-configurationlocal-data-management

FyAgent

Manage models, skills, MCP services, prompts, and AI app settings from one local desktop application.

FollowAgents review · FARS-2.1
Recommended
84/ 100 5-point scale 4.2 / 5
1 2 3 4 5 6
Per-dimension scores and reasoning
1Trust23 / 29 · 4.0/5

The security policy maps local storage, remote data, imports, sync, deep links, proxy traffic, and credential flows. CI uses read-only permissions, disables persisted checkout credentials, pins third-party Actions to commits, and schedules CodeQL. The README promises previews before writes and post-save checks, while the policy requires informed consent for imported executable integrations. Deductions apply because the application operates with the user's filesystem authority, its proxy may bind beyond loopback, and the supplied source does not show every confirmation UI, secret-storage implementation, or network control. Backup and recovery documentation is referenced but not included, so rollback is adequate rather than complete. Project ancestry, CC Switch derivation, authorship, and license attribution are concrete enough for full source-attribution credit.

2Reliability11 / 14 · 3.9/5

The README, package metadata, security policy, CI, and tests consistently describe the product, supported platforms, toolchains, and architecture. CI specifies version-contract, type, formatting, unit, browser, Rust, native credential, and architectural checks. Dependency setup is disciplined through a fixed pnpm version, frozen-lockfile installation, and toolchain verification, but several npm declarations use version ranges and builds depend on external downloads plus mise, Rust, uv, and managed Python. CI aggregates failed-step diagnostics, although the supplied material does not demonstrate end-user error messages or recovery guidance, so failure messaging is not fully evidenced.

3Adaptability16 / 18 · 4.4/5

The intended audience and scenarios are explicit, and Provider, MCP, and Prompt concepts are translated into user-oriented language. The feature matrix names concrete compatibility differences across AI applications, while the security policy thoroughly defines trust boundaries, inclusions, exclusions, and invalidation triggers. Installation and CI cover macOS plus Windows x64 and ARM64. Trigger precision loses a point because scanning, importing, saving, deep-link, and proxy triggers are described without the corresponding implementation needed to confirm that every automatic action is narrowly gated.

4Convention16 / 18 · 4.4/5

The README provides strong navigation, quick-start instructions, a feature table, FAQ entries, community routes, development commands, and multilingual manual links. Naming is stable, with the VibeKey-to-FyAgent transition explained. Licensing clearly separates PolyForm Noncommercial code, inherited CC Switch MIT portions, third-party material, and commercial authorization. Limitations are candid, including active development, latest-release-only security support, unavailable Windows formats, and proxy exposure. Deductions reflect that the full FAQ, installation manual, and release-note contents were not supplied. Maintenance channels and security timelines are clear, but the unknown publisher identity and reliance on a named individual account leave organizational continuity underdocumented.

5Effectiveness10 / 13 · 3.8/5

The documented installation, scan, configuration, preview, and verification workflow is directly usable, and screenshots plus the feature matrix support practical multi-tool management. Centralizing models, Skills, MCP services, prompts, and memory plausibly adds value, but comparative evidence and user outcomes are absent. Local storage, the lack of a project cloud backend, and consolidation improve the benefit side; the noncommercial license, substantial development toolchain, external integrations, and backup burden prevent full cost-benefit credit.

6Verifiability8 / 8 · 5.0/5

Major claims are tied to named files, configuration fields, source locations, version v3.18.0, asset hashes, and test contracts. The README, security policy, package configuration, CI, and architectural tests provide substantial cross-source corroboration. The security policy explicitly separates checkable facts from its renderer scoping decision and lists conditions that invalidate the inference, earning full fact/inference-separation credit. This was not executed, so declared CI and tests are not evidence that this revision actually passed them.

Evidence confidence: Low Reviewed Sep 11, 2026 Reviewed revision 38b4a986e139
Before you use it
  • The local HTTP proxy may bind to a non-loopback interface; restrict its listener, port, and network exposure, and treat every inbound request as untrusted.
  • The application edits other AI tools' configurations and may import executable MCP, script, or deep-link content; inspect the full command, arguments, environment, and change preview before saving or enabling it.
  • API keys, tokens, WebDAV/S3 sync payloads, and shared configurations cross trust boundaries; the supplied material does not fully expose secret-storage, log-redaction, or sync-encryption implementations.
  • The project is still under development and only the latest release receives security updates; back up ~/.fyagent and verify checksums, signing status, and build attestations before upgrading.
  • FyAgent-owned code uses PolyForm Noncommercial rather than an OSI-approved open-source license; commercial use requires separate written authorization, and mixed-license boundaries should be checked carefully.
See the full review method →

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

FyAgent is a local Windows and macOS desktop application for managing a person's AI software and configuration. It scans for QoderWork CN, TRAE Work CN, WorkBuddy, Grok Build, Codex, Claude Code, and OpenCode, exposing installation, update, launch, authentication, and resource-assignment actions where supported. Its management areas cover models and providers, Skills, MCP services, prompts, and memory files for OpenClaw and Hermes. Configuration edits are previewed before writing and checked after saving, while working data is stored locally under ~/.fyagent by default. FyAgent is presented as a configuration control center rather than a hosted agent runtime; its primary output is updated local configuration and memory data for supported AI tools.

FyAgent scans the computer for QoderWork CN, TRAE Work CN, WorkBuddy, Grok Build, Codex, Claude Code, and OpenCode, then shows the installation, update, launch, authentication, or resource-assignment actions available for each detected application. Users can inspect or change model and Provider settings, install Skills from local sources or the discovery page, and assign those Skills to supported software. They can also add, import, manage, and assign MCP services. Prompt management covers Grok Build, Codex, Claude Code, OpenCode, Gemini, OpenClaw, and Hermes. The memory module edits long-term memory files for OpenClaw and Hermes, plus OpenClaw daily memory. Before a configuration is written, FyAgent previews the change scope; after saving, it checks the result and can expose a connection-testing action. Its working data is stored locally in ~/.fyagent by default.

  1. An individual using Codex, Claude Code, and OpenCode who wants one desktop interface for reviewing and changing models, Skills, MCP services, and prompts.
  2. A user unfamiliar with Provider, MCP, or prompt configuration files who needs visible settings and a preview before changes are written.
  3. Someone distributing the same Skills or MCP services across several supported AI applications instead of configuring every application separately.
  4. An OpenClaw or Hermes user maintaining long-term memory files, including OpenClaw's separate daily memory.
  5. A user who prefers configuration data to remain on the local machine under ~/.fyagent rather than living solely in a hosted control panel.

What are this agent's strengths and limitations?

Pros
  • Combines model, Provider, Skills, MCP, prompt, and memory management in one local interface instead of requiring manual edits across several AI applications.
  • Previews configuration changes before writing and checks the result after saving, providing a concrete safeguard against accidental edits.
  • Explicitly covers multiple products and configuration targets, including Codex, Claude Code, OpenCode, Gemini, OpenClaw, and Hermes, rather than tying management to one model provider.
  • Stores working data locally in ~/.fyagent by default and documents configuration locations and backup procedures.
  • Publishes installers for Windows x64, Windows ARM64, and a signed and notarized macOS build.
Limitations
  • The project remains under active development, and users are told to back up important configuration and review the trust notes for every upgrade.
  • Feature coverage differs by AI application; installation, update, launch, authentication, and resource allocation are available only where supported.
  • Only Windows and macOS are documented, with no Linux installation path.
  • Windows users receive an NSIS installer but no MSI or portable ZIP.
  • FyAgent is source-available rather than OSI-defined open source. Its own components and modifications use the PolyForm Noncommercial License 1.0.0, and commercial use requires separate written authorization.

How do you install or deploy this agent?

FyAgent requires Windows or macOS. Download the installer for the current platform from GitHub Releases: FyAgent-X.Y.Z-macOS.dmg for macOS, FyAgent-X.Y.Z-Windows-x64-setup.exe for Windows x64, or FyAgent-X.Y.Z-Windows-arm64-setup.exe for Windows ARM64. Windows releases use NSIS and do not include an MSI or portable ZIP. The macOS build is signed with an Apple Developer ID and notarized by Apple. Before installation, review the release's trust statement and verify its checksum, signing-status.json, and build attestation. No end-user command-line installation or universal credential setup is documented; authentication for individual AI applications is available through FyAgent only where that integration supports it.

How do you use this agent?

Launch FyAgent, open AI Software Configuration, and scan the machine for installed applications. Select a detected application and configure its model and Provider, install or assign Skills, add or import MCP services, manage prompts, or use the available authentication action. Review the proposed scope in the preview before saving. After the write completes, follow the interface prompts to confirm the configuration or test the connection. OpenClaw and Hermes users can use the memory module for long-term memory files, with daily memory management also available for OpenClaw. Back up important configuration before upgrading FyAgent.

How does this agent compare with similar options?

The current FyAgent desktop application evolved from CC Switch and retains the original copyrights and MIT licensing for inherited code. FyAgent's own components and modifications use the PolyForm Noncommercial License 1.0.0. The supplied material does not include a feature-by-feature comparison, so it does not establish migration compatibility or relative coverage.

FAQ

Does FyAgent upload my data to the cloud?
Working data is stored locally in ~/.fyagent by default. The supplied material does not document telemetry, synchronization, or other upload behavior, nor does it fully describe the data flow for model connections, authentication, or the discovery page.
Can I use FyAgent commercially?
FyAgent-owned components and modifications are licensed under PolyForm Noncommercial License 1.0.0, so commercial use requires separate written authorization. Code inherited from CC Switch remains under the MIT license.
Does FyAgent run AI agents itself?
Its documented role is to manage models, Providers, Skills, MCP services, prompts, authentication, and memory configuration for local AI software. It is not described as an independent task-execution service or hosted agent runtime.
What information should I provide when reporting an installation or configuration problem?
After checking the troubleshooting material, provide the FyAgent version, operating system, affected AI application, and steps already attempted in the Q&A forum. A consistently reproducible software defect can be filed through the Bug Report template.
Does a NotSigned result prove that a release file is unsafe?
No. NotSigned describes signing status only and is not sufficient evidence of safety. Users are also instructed to verify checksums, signing-status.json, build attestations, and the release notes.

Related agents