ctxjev-mcp 0.1.12 → 0.3.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +13 -1
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -17,6 +17,14 @@
|
|
|
17
17
|
`ctxjev-mcp` speaks plain stdio MCP: the same binary works with every host below, and only the
|
|
18
18
|
config shape differs.
|
|
19
19
|
|
|
20
|
+
> **Before you expect it to save tokens:** an MCP tool returns data to whoever called it; it can't
|
|
21
|
+
> remove anything from the host's own context. And to score its history, the agent has to send
|
|
22
|
+
> that history as tool arguments, which the host model pays for in its own output tokens. It's
|
|
23
|
+
> useful when something acts on the scores (an agent framework that manages its own context), not
|
|
24
|
+
> as a drop-in token saver for Claude Code, Codex, or Copilot. For Claude Code, the
|
|
25
|
+
> [ctxjev plugin](https://github.com/x96x64/ctxjev#the-claude-code-plugin) is the integration that
|
|
26
|
+
> actually helps.
|
|
27
|
+
|
|
20
28
|
## Setup
|
|
21
29
|
|
|
22
30
|
Get a key at [console.typesafe.ai/settings/keys](https://console.typesafe.ai/settings/keys) (no
|
|
@@ -59,6 +67,10 @@ codex mcp add ctxjev --env TYPESAFE_API_KEY=... -- npx ctxjev-mcp
|
|
|
59
67
|
|
|
60
68
|
`entries` is `{ id, role: 'user'|'assistant'|'tool', toolName?, content, timestamp }[]`.
|
|
61
69
|
|
|
70
|
+
Entry content and the goal are sent to TypeSafe AI's Jev API. Common secret formats (API keys,
|
|
71
|
+
tokens, private-key blocks, `NAME=value` credentials) are masked to `[REDACTED]` first, on a
|
|
72
|
+
best-effort basis.
|
|
73
|
+
|
|
62
74
|
## Example Response
|
|
63
75
|
|
|
64
76
|
Calling `prune_history` with two entries, one obviously relevant to the goal and one not, returns:
|
|
@@ -71,7 +83,7 @@ Calling `prune_history` with two entries, one obviously relevant to the goal and
|
|
|
71
83
|
],
|
|
72
84
|
"savings": {
|
|
73
85
|
"totalEntries": 2, "keptEntries": 1, "droppedEntries": 1, "summarizedEntries": 0,
|
|
74
|
-
"totalTokens": 13, "
|
|
86
|
+
"totalTokens": 13, "droppedTokens": 5, "summarizableTokens": 0
|
|
75
87
|
},
|
|
76
88
|
"usage": { "inputTokens": 406, "outputTokens": 36 }
|
|
77
89
|
}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "ctxjev-mcp",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.3.0",
|
|
4
4
|
"description": "MCP server exposing ctxjev's Jev-based context scoring/pruning as tools for any MCP-capable agent host.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"mcp",
|
|
@@ -40,7 +40,7 @@
|
|
|
40
40
|
"dependencies": {
|
|
41
41
|
"@modelcontextprotocol/sdk": "^1.30.0",
|
|
42
42
|
"zod": "^4.6.5",
|
|
43
|
-
"ctxjev-core": "0.
|
|
43
|
+
"ctxjev-core": "0.3.0"
|
|
44
44
|
},
|
|
45
45
|
"devDependencies": {
|
|
46
46
|
"@types/node": "^22",
|