Cost calculator for Vercel AI SDK token usage. Supports OpenAI, Anthropic, Google, xAI (Grok), and DeepSeek with long context pricing, prompt caching, and reasoning tokens.

Conclave AI Gemini agent — wraps @google/genai with responseSchema structured output. Long-context slot per decision #10.

Never lose context again. Long Context Manager (LCM) compresses Claude Code sessions into searchable memory — every message preserved, every insight remembered.

Build an agent session once, then fork it. A **master session** is a long context you pay to build a single time — exploration, documentation, domain knowledge. Every **child session** forks that context instead of rebuilding it.

OpenClaw enhanced LanceDB memory plugin with hybrid retrieval (Vector + BM25), cross-encoder rerank, multi-scope isolation, long-context chunking, and management CLI

Toggle OpenAI GPT-5.6 and GPT-6 models between pi's 272K context window and 1.05M tokens.

Up-to-date LLM pricing reference and cost calculator — Claude, GPT, Gemini, Grok, Mistral, DeepSeek, Groq and more. Long-context tiers, SDK usage objects, spend tracking. Zero dependencies, library + CLI.

OpenCode plugin to add long-context Claude and GPT-5 variants via GitHub Copilot

Extension for the pi coding agent: enforce a context token budget (defaulting to the model's configured context window; 200k when unknown) by forcing compaction, including mid-turn during long tool loops, instead of riding long-context models up to their

Evaluation runner for wasmagent — multi-model multi-axis benchmark harness with paired statistics + Pareto report. Targets the gaps in evomerge-style single-axis evaluation: long-context recall, multi-turn memory, agent trajectory, latency, cost.

Recursive Language Model as an MCP server: recursive long-context Q&A and a persistent Python kernel.

Recursive Language Model and Continual Harness in one MCP server: long-context Q&A, a Python kernel, and durable memory.

Extension for the pi coding agent: enforce a context token budget (default 200k) by forcing compaction, including mid-turn during long tool loops, instead of riding long-context models up to 1M.

Implements the WebIDL algorithms for converting to and from JavaScript values

Public logs API for OpenTelemetry

Cap model context windows so pi's built-in auto-compaction fires earlier. Zero-config 200k default for long-context Claude models; configurable for anything else.

A polyfill for the `interestfor` API

React hook for detecting click, tap or point and hold event. Easy to use, highly customizable options, thoroughly tested.

Context-compression proxy that lets AI coding agents run for days — billions of tokens through one context window. Sits between any agent (Claude Code, Codex, Cursor, Aider, …) and its model API, folding consumed conversation into reversible, prefix-cache

A Long class for representing a 64-bit two's-complement integer value.