OpenCode Context Optimizer (OCCE)
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| code_skeletonA | Extract an AST outline of classes, functions, signatures, argument type hints, and docstrings from a source file, replacing bodies with '...'. Consumes 80-94% fewer tokens than raw file reading. |
| symbol_sliceB | Extract the exact implementation of a single class or function by symbol name (e.g., 'PaymentGateway' or 'PaymentGateway.charge' or 'charge'), avoiding the need to read the entire file. |
| smart_execB | Execute a shell command with real-time ANSI escape stripping, carriage-return collapse, package manager progress bar & spinner suppression, and intelligent head/tail truncation. |
| test_compressA | Run test suites (pytest, cargo test, go test, jest/vitest) and output only failing tests, assertion errors, and clean stack frames, collapsing passed tests into a 1-line summary. |
| diff_compactA | Generate a zero-context (-U0) git diff of uncommitted changes or specific files, eliminating unchanged surrounding context lines and git index boilerplate. |
| token_ledgerB | Check current session token usage, cumulative savings achieved by OCCE, and free-model attention health zone (GREEN / YELLOW / RED alert). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool has a distinct role: AST outline vs. single-symbol extraction vs. command execution vs. test output vs. diff vs. token accounting. The only mild overlap is code_skeleton vs. symbol_slice, where both read source code, but the outline-vs-implementation distinction in the descriptions resolves it.
All names are snake_case two-word compounds in a modifier+noun style (code_skeleton, symbol_slice, test_compress, diff_compact, token_ledger), with smart_exec being the only slight deviation in verb-vs-noun form. Readable and predictable overall.
Six tools for a context-optimization server is well-scoped: two for reading code, three for compressing execution/test/diff output, one for telemetry. No filler tools.
The surface covers code reading, command execution, test output, diffs, and token accounting, which are the core levers of context reduction. A small gap is the absence of a symbol/search tool to locate symbols across files before calling symbol_slice.