I have sat next to the mirror-phone wall at 14:12 while MCP host was this close to calling the wrong MCP tool.
Not a model failure. A posture failure.
MCP creator 2026: clients already know progressive discovery / tool search. Most hosts still dump.
Someone on the floor said, out loud, do not naively dump 100 tools into the window. MCP host had started earlier. The tool list was frozen in config.toml / .mcp.json / a skill file. The only “safe” move anyone trusted was:
- Kill MCP host (or its MCP child)
- Edit the server list
- Restart the host
- Lose the plan the model already paid for
That is the ceremony. MCP default dumps all schemas; Skills metadata is a few dozen tokens until triggered
The Aha: tool-search=on is posture. The host defers schemas without a new MCP binary. Real-time nested state is now a primitive — as native as a hashmap — via Kiponos.io. The host stays up. The leaf moves. The next tool call already obeys.
The problem: Progressive disclosure as a live switch, not a client rewrite
Agentic programming in 2026 has a public fight: MCP vs CLI vs Skills. The likes on MCP and Grok Build posts are the tell. The industry is stuck on how to expose tools.
The missing primitive is live nested state:
| Where the roster hid | What you restart | What you lose |
|---|---|---|
| MCP schema dump at session start | The host, sometimes the child npx
|
16–72% of the window before the user speaks |
~/.grok/config.toml / Cursor mcp.json
|
MCP host | The open turn |
| SKILL.md / AGENTS.md | The next turn, sometimes the host | Context you already bought |
| “Just reconnect the server” | The MCP session | Tool-space you finally had right |
MCP default dumps all schemas; Skills metadata is a few dozen tokens until triggered
That is not a protocol bug you wait out. That is a process that cannot get() a tree that moved.
What teams believe
| Belief | Production |
|---|---|
| Feature flags cover MCP write | Flags are another product, another delay |
| Restart the agent host — it is cheap | Cheap until 14:12 ate the open turn |
| MCP is the standard so we are done | MCP is how you call. It is not live posture |
| Skills will progressive-disclose it | Skills are files. Files do not fan out |
What is Kiponos here
Kiponos is a live nested config hub. Dashboard edit → WebSocket delta → in-memory tree on every connected process. Hot-path get() is a local map lookup. One WebSocket per process lifetime.
It is not an MCP server. It is not a Skill. It is the control plane those things do not ship: which tools may run, how fat a result may be, whether execute is allowed, whether a child inherits posture — this turn.
Profile shape: ['my-app']['v1.0.0']['dev']['base']. The leaf in this story is tool-search under examples/agentic-mcp-0910-pm-progressive/tool-search.
Architecture
The model still calls tools. The wrapper reads tool-search before the call. No schema dump policy in argv.
Config tree
examples/agentic-mcp-0910-pm-progressive:
tool-search: off
schema-budget: "4000"
result-cap: "2000"
tool-search: "on"
execute-gate: "plan"
Eight-plus keys when you pair tool-search with sister dials (budget, mute, pause). They move together on the mirror-phone wall.
Integration
Java tool wrapper (MCP host just calls it):
import io.kiponos.Kiponos;
import io.kiponos.Folder;
Folder gate = Kiponos.createForCurrentTeam()
.getFolder("examples/agentic-mcp-0910-pm-progressive");
String roster = String.valueOf(gate.get("tool-search"));
// MCP host tool: refuse the dangerous MCP call when posture moved
Python peer:
from kiponos import Kiponos
k = Kiponos.connect(quiet=True)
try:
posture = k.get("examples/agentic-mcp-0910-pm-progressive/tool-search", "live")
if not str(posture):
raise PermissionError("Progressive disclosure as a live switch, not a client rewrite gated live — host not restarted")
finally:
k.disconnect()
React/Angular: Node BFF @kiponos/react/server / @kiponos/angular/server. SPA never holds Connect tokens.
Runnable mesh: https://github.com/kiponos-io/kiponos-io/tree/master/examples/java/agentic-mcp-0910-pm-progressive — ./gradlew test run.
Real scenarios
| Event | Without Kiponos | With Kiponos |
|---|---|---|
| Three MCP servers at boot | 72% of the window gone | Live roster / schema-budget; dump shrinks |
| User says “no writes” mid-turn | Restart MCP host | Set tool-search; next call already denies |
| Second host started earlier | Paste the json into chat | Both get() the same leaf |
| Subagent spawned | Frozen argv copy | Child get()s inherit-posture
|
| Tool returns 500k tokens | Context rot, doom loop |
result-cap in the wrapper |
Performance (this path)
- Wrapper
get()is in-process after bootstrap. - One WebSocket per MCP host lifetime — not per MCP call.
- A dashboard edit is a delta of
tool-search, not a config.toml reload. - You do not spend model tokens to “please restart MCP host.”
- Schema tax is optional once the live set is small.
Compare to alternatives
| Approach | Honest fit | Why it still restarts |
|---|---|---|
| More MCP servers | More integrations | More schema, worse selection |
| Skills progressive disclosure | Good docs, lazy load | Not a multi-host bus |
| Claude Code tool search | Better client | Other hosts still dump |
| Redis poll in the tool | Shared, extra RTT | You invented a hub with worse UX |
| Feature-flag SaaS | Product experiments | Rarely session-safe for MCP |
| Kill -9 the npx child | Janitor | The turn still died |
When not to use Kiponos
| Situation | Why |
|---|---|
| Tool schema itself changed (new argument) | That is a code/MCP host restart |
| Secret rotation of Connect / MCP OAuth tokens | Credentials are not live knobs |
| One-off local script, no peers | A hub is overkill |
| Browser-only “SDK in the SPA” | Forbidden — tokens leak or defaults lie |
Pair tool-search with a sister dial
tool-search rarely moves alone on the mirror-phone wall. Pair it with schema-budget, result-cap, or execute-gate so you do not fix Progressive disclosure as a live switch, not a client rewrite by inventing a second incident.
Why MCP host is the wrong restart target
MCP host is good at calling tools. It is not a control plane. Killing it to flip tool-search teaches the on-call that judgment requires a process ID. The mirror-phone wall already disagrees.
Getting started (15 minutes)
- TeamPro on kiponos.io → Connect →
KIPONOS_ID/KIPONOS_ACCESS/ profile['my-app']['v1.0.0']['dev']['base']. - Clone github.com/kiponos-io/kiponos-io.
cd examples/java/agentic-mcp-0910-pm-progressive && cp kiponos.local.env.example kiponos.local.env-
./gradlew test run— printsexamples/agentic-mcp-0910-pm-progressive/tool-search=... - In the dashboard, change
tool-search. Keep MCP host up. No MCP reconnect. - Point the tool wrapper at the same leaf. Do not ship a new server binary to flip Progressive disclosure as a live switch, not a client rewrite.
Further reading
The moral
If flipping Progressive disclosure as a live switch, not a client rewrite requires restarting MCP host or reconnecting MCP, you do not have agentic control. You have a hope with a process ID.
MCP, Skills, and tools are how agents act. Kiponos is the live primitive they do not ship — so the mirror-phone wall can change its mind without killing the session.
How to try: examples/java/agentic-mcp-0910-pm-progressive and ./gradlew test.
Top comments (0)