External MCP plugins (UUU/BCU, U-Boot) for the FC MCP server
Listed on its own since 2026-09-27. Until then it was part of fc-mcp, where its earlier checks remain.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"fc-mcp-external","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
What is FC? ----------- FC (Framework Coordinator), is an open source coordinator service among different embeded frameworks. It's built using the same principles as `Mesos <http://mesos.apache.org/>`_, provides different frameworks (e.g., LAVA, Labgrid) with API's for resource management and scheduling across entire board farms. Background ---------- There are many different systems in embedded test community, see `this <https://elinux.org/Test_Systems>`_. They are good at different scenarios, e.g. * **Test automation:** LAVA, Fuego, autotest, avocado, TI VATF, U-Boot test tool, CI-RT R4D, Baylibe Lab in a Box...... * **Development automation:** Labgrid...... Let's take a look at why FC is useful by going back in time. .. image:: doc/images/era.svg * **Traditional deployment era:** Early on, as nearly all above systems require dedicated control of the board resources, organizations have to afford multiple series of hardware to meet the requirements of different frameworks, which leads to resource waste when leverage different systems. * **FC deployment era:** As a solution, FC was introduced. It allows different frameworks could share same series of hardware, move resources between systems seamlessly. Different systems will continue work without aware existence of other systems. So, FC could be treat as "scheduler's scheduler". The initial idea comes from `Jan's proposal to move boards between LAVA and labgrid on ATS2019 <https://elinux.org/Automated_Testing_Summit_2019>`_, we realized the FC based on that thought (Thanks community to inspire the idea). Compared to other resource management system, it has low invasion to frameworks. This means: unlike mesos which you should write your mesos scheduler for your framework & inject it into your framework code, FC won't change your framework code, it tries to adapt to your framework. But still, your framework need next two features: * Your framework needs to have a job queue which FC could monit…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.