Author here. Started building this in January. The protocol went public
on GitHub in February; the companion paper landed on Zenodo in March
(https://doi.org/10.5281/zenodo.19040913). Today the whole stack is
finally installable from every registry — 17 packages across npm, PyPI,
crates.io, and Go modules.
The thesis: agents fail from missing infrastructure, not missing
intelligence. Human societies didn't coordinate by perfecting individuals —
they invented institutions. DCP-AI is an attempt at that scaffolding for
autonomous AI: cryptographic identity bound to a responsible human,
machine-readable intents with policy gating, tamper-evident audit chains,
a formal lifecycle/succession/rights framework, and an agent-to-agent
arbitration protocol.
Hybrid post-quantum crypto (Ed25519 + ML-DSA-65 composite, NIST FIPS
203/204/205 conformant). Apache-2.0.
Quick things to try:
- Interactive playground: https://dcp-ai.org/playground/
- npm create @dcp-ai/langchain my-app (or /crewai, /openai, /express)
- Docs: https://docs.dcp-ai.org
Happy to discuss design choices. Most interested in pushback on DCP-07
(agent-to-agent conflict arbitration) — that's where the design was
hardest to commit to.
The ast-grep + ripgrep combination for semantic context is the right architectural choice. Pure embedding-based retrieval tends to fail on codebases with non-trivial inheritance hierarchies or polymorphism, where structural search beats semantic similarity. I'd be curious how you're balancing the two: does ast-grep run first as a structural filter, with ripgrep for content matching, or are they used independently depending on the query type?
On the multi-provider abstraction: Anthropic, OpenAI, and Gemini have meaningfully different tool-calling schemas, and Codex (the CLI tool) adds another layer because it wraps OpenAI's API but with its own conventions. How are you handling the schema translation? Most "multi-provider" implementations I've seen end up with provider-specific code paths that defeat the abstraction.
ACP support is interesting. I haven't seen many agents implement it yet, mostly MCP. Is your read that ACP is going to gain adoption, or is including both more about hedging?
The local inference angle (LM Studio, Ollama) matters for use cases where source code can't leave the network. Have you benchmarked which open models hold up reasonably for tool-calling-heavy workflows? In my experience most local models below 70B struggle with multi-turn tool use even when their raw code generation is decent.
Rust + Ratatui is a strong DX choice. Will check out the DeepWiki.