Repository Issue Activity (beta)

#960 Support more AI providers

#959 Default OpenAI model gpt-5-mini is scheduled for shutdown on 2026-12-11

#951 How to feed shell command to agent context?

#955 Feature Suggestion: Optional token metering & paid API key support via `neuforge-pay`

#903 Marketplace entry name is used as an install directory name without containment, escaping the managed root

#88 Chain Agent only performs first generation when cumulative=True

#930 Feature request: official HOL Guard before_tool_call security example

#908 ACP registry CI failures

#918 Documented install (uv tool install -U fast-agent-mcp) sticks at 0.9.30: 0.10.x pins a fastmcp-slim pre-release

#922 Feature Suggestion: Optional token metering & paid API key support via `neuforge-pay`

#766 Add pluggable tool-search for large MCP toolsets

#210 Support for AG-UI protocol

#889 AttributeError: 'int' object has no attribute 'casefold' in _is_fatal_retry_error when provider returns numeric error code

#847 Calfskin wallet for PRs?

#849 Tool selection when multiple MCP servers expose overlapping capabilities

#836 Relaxed `pydantic` dependency

#837 feat(run): add --timeout flag for bounded execution

#781 `tools` must not be an empty array

#827 ACP initialize does not advertise mcpCapabilities (http/sse)

#811 Two defense-in-depth hardenings: tar extraction filter + Windows shell=True in file-open

#127 feature: add support for LiteLLM Provider

#774 Agent identity verification across MCP server boundaries

#273 Allow registration of custom LLM provider classes

#599 Question regarding capabilities negotation

#600 change the client to send the elicitation/create result immediately on SSE receipt, independent of the tools/call HTTP lifecycle