projects
Why these tools exist
Writing code got cheap. Checking that it holds up did not.
An agent steering for a green test suite gets a green test suite. An assertion comparing a value with itself passes exactly as hard as one that proves something, and so does a stub for a method that stopped existing months ago. A suite like that is a signature under a document nobody read.
Outside the tests it goes the same way. An agent that does not know your architecture guesses at it, and a guess that is wrong looks just like one that is right. An agent allowed near your database is allowed everything there, unless somebody drew that line.
That is what these tools are for. Not producing code faster, but seeing what the code you already have actually does.
Where it goes wrong
- Your suite is green, but you do not know what it would catch.
- Chaos-MCP
- Your agent needs your database and must not have the run of it.
- Argos-MCP
- Your agent guesses at your architecture, every session again.
- Knossos-MCP
- Your tests pass, and prove nothing.
- Momus-MCP
- You do not know what you are pulling in when you install an MCP server.
- MCP Observatory
- You do not know where a session's time goes.
- Claude Timestamp
How they work together
before you install
Knowing what an MCP server brings in with it.
before the agent starts
It has to know what depends on what, without rereading the repository.
while it builds
Reaching the data without being able to change it, and seeing where the time goes.
before you merge
Proving that green is also healthy.
All six
| tool | work | when | status |
|---|---|---|---|
| Chaos-MCP | test quality | before you merge | public |
| Argos-MCP | agent tooling | while it builds | public |
| Knossos-MCP | architecture | before the agent starts | public |
| Momus-MCP | test quality | before you merge | public |
| MCP Observatory | ecosystem | before you install | public |
| Claude Timestamp | agent tooling | while it builds | public |