SignalMD - 2026 Pricing, Features, Reviews & Alternatives

SignalMD is a cloud-based call analysis platform built for teams developing and operating Retell voice agents. It reads production calls automatically, groups repeated failures by the change that resolves them, and pins each issue to the exact step, tool, or transition in the agent flow where it occurs. No metric configuration or judge prompts are required; a one-page brief is drafted from the agent's existing configuration and approved by the user before analysis begins. SignalMD supports Conversation Flow, multi-state, and single-prompt Retell agent types, and connects via Retell API key with no webhook changes needed. Deployment is fully cloud-based via a web app, REST API, and MCP server, with infrastructure hosted in US East.

SignalMD covers the full fix lifecycle: a Copilot side panel reads the issue board, answers questions about why a failure occurs using the calls behind it, drafts changes to prompts or agent flow, and runs a quick pre-test simulation before formal testing begins. Changes are written to a draft and tested against a copy of the live agent via Retell simulation; publishing requires a test on the draft's exact text and supports rollback. After a new version accumulates calls over a configurable grace period (default 7 days), each issue receives a verdict based on statistically meaningful differences between versions, not simulations. A flow view displays every step alongside the calls that took it, colored by outcome, and a call-level transcript view marks each problem at the moment it occurred.

SignalMD's pricing model is credit-based with flat monthly plans starting at $49, scaling to enterprise tiers. No per-metric charges apply, and exhausted credits result in calls being queued rather than deleted. The REST API supports reading issues, versions, and test results from scripts or CI pipelines, with keys scoped to a workspace or individual agent. An MCP server is also available alongside the app and API. The tool runs its own model reads, requiring no external model key; if model reads are unavailable or exceed a plan cap, analysis continues and affected calls are flagged on screen.