I dug into this further. Tested empirically and read the code.
Confirmed: context-mode cannot intercept MCP tool responses. The PreToolUse hook (hooks/pretooluse.sh) matches only Bash|Read|Grep|Glob|WebFetch|WebSearch|Task. When I called my obsidian MCP's obsidian_list via MCP, the response went straight into context — zero entries in context-mode's FTS5 database. The web fetches from the same session were all indexed.
The context-mode skill (SKILL.md) actually acknowledges this at lines 71-77 with an "after-the-fact" decision tree for MCP output: if it's already in context, use it directly; if you need to search it again, save to file then index. But that's damage control — the context is already consumed. You can't un-eat those tokens.
The architectural reason: MCP tool responses flow via JSON-RPC directly to the model. There's no PostToolUse hook in Claude Code that could modify or compress a response before it enters context. And you can't call MCP tools from inside a subprocess, so the "run it in a sandbox" pattern doesn't apply.
So the 98% savings are real but scoped to built-in tools and CLI wrappers (curl, gh, kubectl, etc.) — anything replicable in a subprocess. For third-party MCP tools with unique capabilities (Excalidraw rendering, calendar APIs, Obsidian vault access), the MCP author has to apply context-mode's concepts server-side: return compact summaries, store full output queryably, expose drill-down tools. Which is essentially what you suggested above.
Still very high value for the built-in tool side. Just want the boundary to be clear.
Really intrigued and def will try, thanks for this.
In connecting the dots (and help me make sure I'm connecting them correctly), context-mode _does not address MCP context usage at all_, correct? You are instead suggesting we refactor or eliminate MCP tools, or apply concepts similar to context_mode in our MCPs where possible?
Context-mode is still very high value, even if the answer is "no," just want to make sure I understand. Also interested in your thoughts about the above.
I write a number of MCPs that work across all Claude surfaces; so the usual "CLI!" isn't as viable an answer (though with code execution it sometimes can be) ...
I shared a lot of these uninformed (speaking only for myself) opinions, but have a lot more respect for Peter after listening to his recent 3hr interview with Lex Fridman. I came away liking and respecting.
Slightly off topic but I am writing an Excalidraw MCP that allows Claude Code and Claude.ai to create Excal drawings and then iterate on them — I gave Claude the ability to “see” the resulting drawing via a tool that runs a rendering pipeline and returns a png to the model. It’s producing the Olympic logo pretty nicely ;-)
I strongly recommend against Time Machine to NAS — just mentioned it in an article today as “works great, until it inevitably and catastrophically fails”.
Reliable for me is Kopia from Mac to S3 compatible volume (minio) on Synology.
Am I the only one who thinks (mobile safari with an ad blocker) it’s a jackass move to show a fake popup that leads to “must be an adblocker” and prevents page viewing? Or maybe just bad JS junk causing it?
Update: OMG I turned ad blocker off and what a disaster of a site.
For anyone using Qwen3-VL: where are you running it? I had tons of reliability problems with Qwen3-VL inference providers on OpenRouter — based on uptime graphs I wasn’t alone. But when it worked, Qwen3-VL was pack-leading good at AI Vision stuff.
Ha, I have been steeling myself for a long chat with Claude about “how the F to get AI Studio up and working.” With paying being one of the hardest parts.
Without a doubt one essential ingredient will be, “you need a Google Project to do that.” Oh, and it will also definitely require me to Manage My Google Account.
80% of MCPs are thin wrappers over APIs . Yes they stink.
A well written remote OAuth MCP need not stink. Tons of advantages starting with strong security baked in.
I like Cloudflare Code Mode as an MCP pattern. Two tools, search and execute.
1M Opus 4.6 also reduces the penalties of MCP’s context approach. Along with tool search etc.