agentrt
Airymax AgentRT Transcend context limits. Achieve near-infinite memory.
openairymax/agentrt is ranked 10,951 with 1,254 stars.
openairymax/agentrt is ranked 10,951 among these 8 peers; chunxiaoxx/nautilus-compass is ranked 10,949.
| Name | Rank | Metrics |
|---|---|---|
| hashgraph-online/standards-sdk | 10,919 |
|
| plurai-ai/intellagent | 10,925 |
|
| ChristopherKahler/paul | 10,947 |
|
| chunxiaoxx/nautilus-compass | 10,949 |
|
| RUC-NLPIR/Search-o1 | 10,960 |
|
| hexsum/Mojo-Weixin | 10,963 |
|
| Gentleman-Programming/agent-teams-lite | 10,974 |
|
| firecrawl/web-agent | 10,976 |
|
| Parameter | Value |
|---|---|
| Language | C |
| Category | AI Engineering |
| Subcategory | Give agent knowledge, Build agent workflow (SDK) |
| Keywords | Agent-ready APIs, Context engineering, Memory management, Planning |
| Stars | 1,254 |
| Forks | 109 |
| Contributors | 1 |
| Created date | 2026-03-09 |
| Updated date | 2026-10-16 |
| Stars in 1 day | 0 |
| Stars in 7 days | -2 |
| Rank | 10,951 |
| Previous rank | 10,914 |
| Rank change | -37 |
| 7-day stars rank | 18,330 |
| Previous 7-day stars rank | 18,279 |
| 7-day stars rank change | -51 |
| Total stars rank | 9,354 |
| Previous total stars rank | 9,347 |
| Total stars rank change | -7 |
It provides the mechanisms an agent team needs in order to run for a long time on real hardware: a micro-kernel core with scheduling, syscalls and memory primitives, a three-stage cognitive loop (cognition → execution → memory), layered session and cross-session memory, a four-layer security dome that fails closed by default, an IPC and agent-to-agent protocol stack, a single gateway process that speaks HTTP / WebSocket / SSE / MCP / A2A / OpenAI and translates to JSON-RPC 2. 0
, a fleet of long-running daemons that orchestrate all of the above, a command-line interface and a terminal UI for interactive use. Everything is written in C11 with language bindings for Python, Go, Rust and TypeScript. The runtime is designed so that a single laptop and a multi-agent deployment use exactly the same programming model. This repository is the **release aggregate** for the runtime. The individual components are maintained in their own repositories and are wired in as
git submodules, pinned to the exact commit that shipped with each release, so that any published version can be reproduced commit for commit. Installation Linux and macOS One command, no compilation required. The installer detects the platform, downloads the matching prebuilt package, and verifies it: The script is shipped with each release. When the downloaded script is saved to a file and executed from disk, the installer self-updates to the newest version. Windows A native PowerShell installer is published
for x86-64 and x86-32: npm A thin wrapper package is published for Node. js users: The package carries no install logic of its own: it locates an existing install first, and otherwise delegates to the same / release assets above (Windows hand-off uses PowerShell). Options Append flags after (or pass them directly when running the script from a file): is meaningful only for the **first bootstrap install**: at that point there is no install root to infer from
, so you must supply it. If several copies are present on one machine, shows which one your shell resolves. Architectures not listed above are not part of the prebuilt matrix; build them from source. First run Start the runtime and its services by launching the terminal UI: reports or overrides the hardware-derived runtime profile ( , or ). The installer picks the profile for you, and keeps watching: if you add RAM or a GPU later, the features that were trimmed are
restored automatically. On Windows the same commands are available through . The same instruction automatically enters a plan-then-execute loop when the job is large. | | **Multi-agent collaboration** | One lead agent orchestrates several child agents, splitting work in parallel and merging results back. Agents interoperate over a documented protocol. | | **Planning and reflection** | Ambiguous goals are clarified interactively, then compiled into an executable task graph that is scheduled. Execution self-reflects and re-dispatches on failure. | | **Persistent memory** | A built-in
free memory backend with L1 session-level and L2 cross-session tiers, supporting retrieval and forgetting. Pluggable commercial backends are
How many stars and forks are recorded?
openairymax/agentrt has 1,254 stars and 109 forks.
What is the repository rank?
openairymax/agentrt is ranked 10,951.
What language and category are recorded?
openairymax/agentrt has 1,254 stars, uses C and belongs to AI Engineering.
Which repository is closest in rank?
openairymax/agentrt is compared with chunxiaoxx/nautilus-compass, ranked 10,949 with 1,254 stars.