FREE PRODUCTION SESSION CAPTURE

Evaluate what your agents actually do, not just what they say.

Evaluate what your agents actually do, not just what they say.

Evaluate what your agents actually do, not just what they say.

End-to-end production capture for AI agents
Softprobe artwork
Softprobe AI agent performance review interface preview
Softprobe AI agent performance review interface preview
Works with OpenCode, LangChain, Vercel AI SDK, Claude Code, Cursor, and Codex
Works with OpenCode, LangChain, Vercel AI SDK, Claude Code, Cursor, and Codex

Why Softprobe

A successful run can still hide a bad decision.

Capture your production agent sessions for free. Inspect each step to find mistakes that don’t trigger errors.

01 · CAPTURE

Capture your agent sessions for free

Copy a single setup prompt into your coding agent to instrument OpenCode, LangChain, or Vercel AI SDK in under two minutes. Stream unlimited production Sessions, including the backend service and third-party API payloads your agent interacts with, with 365-day history at $0.

Review every step

02 · INSPECT

Inspect the full trajectory step by step

Follow every Session in order across user messages, reasoning, model calls, and tool executions, or switch to the end-to-end call graph to see downstream backend and third-party API requests and responses. Open any Step to inspect the exact input, output, timing, and context the agent acted on.

Diagnose what repeats

03 · FIND

Find agent mistakes

Pin silent failures and rule deviations directly to the exact Session and Step where behavior drifted. In Teams Early Access, Softprobe also proposes behavior rules from recorded runs and checks new Sessions automatically.

Prove the improvement

04 · FIX

Fix them with your coding agent

Generate a structured Fix Brief from any Finding in one click. Copy or download the failing Session context, pinned Step evidence, and expected behavior straight into Claude Code, Cursor, or Codex so your coding agent can reproduce and fix the mistake in your repo.

01 · CAPTURE

Capture your agent sessions for free

Copy a single setup prompt into your coding agent to instrument OpenCode, LangChain, or Vercel AI SDK in under two minutes. Stream unlimited production Sessions, including the backend service and third-party API payloads your agent interacts with, with 365-day history at $0.

Review every step

02 · INSPECT

Inspect the full trajectory step by step

Follow every Session in order across user messages, reasoning, model calls, and tool executions, or switch to the end-to-end call graph to see downstream backend and third-party API requests and responses. Open any Step to inspect the exact input, output, timing, and context the agent acted on.

Diagnose what repeats

03 · FIND

Find agent mistakes

Pin silent failures and rule deviations directly to the exact Session and Step where behavior drifted. In Teams Early Access, Softprobe also proposes behavior rules from recorded runs and checks new Sessions automatically.

Verify the improvement

04 · FIX

Fix them with your coding agent

Generate a structured Fix Brief from any Finding in one click. Copy or download the failing Session context, pinned Step evidence, and expected behavior straight into Claude Code, Cursor, or Codex so your coding agent can reproduce and fix the mistake in your repo.

Inside Softprobe Explorer

Built for full-stack agent behavior, not just LLM spans.

Built for full-stack agent behavior, not just LLM spans.

Built for full-stack agent behavior, not just LLM spans.

Go beyond prompt logs with end-to-end backend call graphs, plain-English policies, and direct coding-agent handoff.

01 · Correlate agent steps with backend APIs

Standard LLM trace viewers stop at model prompts and tool return strings. Softprobe links every reasoning step and tool call to the downstream backend services and third-party HTTP payloads your agent triggered.

02 · Turn plain-English policies into automated Findings

03 · Fix in your repo with Fix Briefs and Softprobe MCP

Softprobe Explorer Session Graph linking agent tool calls to downstream backend HTTP payloads

01 · Correlate agent steps with backend APIs

Standard LLM trace viewers stop at model prompts and tool return strings. Softprobe links every reasoning step and tool call to the downstream backend services and third-party HTTP payloads your agent triggered.

02 · Turn plain-English policies into automated Findings

03 · Fix in your repo with Fix Briefs and Softprobe MCP

Softprobe Explorer Session Graph linking agent tool calls to downstream backend HTTP payloads

01 · Correlate agent steps with backend APIs

Standard LLM trace viewers stop at model prompts and tool return strings. Softprobe links every reasoning step and tool call to the downstream backend services and third-party HTTP payloads your agent triggered.

02 · Turn plain-English policies into automated Findings

03 · Fix in your repo with Fix Briefs and Softprobe MCP

Softprobe Explorer Session Graph linking agent tool calls to downstream backend HTTP payloads

Pricing

Pricing

Unlimited data ingestion. Free forever.

Unlimited data ingestion. Free forever.

Unlimited data ingestion.

Free forever.

Long-horizon agents generate megabytes of prompts, reasoning, tool payloads, and backend API responses. Softprobe lets you capture 100% of your production sessions with 365-day history at $0, so you only pay when you want automated QA reviews across your team.

Long-horizon agents generate megabytes of prompts, reasoning, tool payloads, and backend API responses. Softprobe lets you capture 100% of your production sessions with 365-day history at $0, so you only pay when you want automated QA reviews across your team.

Developers

$0

Free forever

Capture, inspect, and debug your production agent sessions at no cost.

Unlimited session ingestion

OpenCode, LangChain & Vercel AI SDK

1-prompt coding-agent setup

Session Explorer (Trajectory, Evidence & Graph)

Manual Findings & 1-click Fix Brief export

Softprobe MCP server (explorer.softprobe.ai/mcp)

365-day Session history

Teams

EARLY ACCESS

$499

per month, billed monthly

Turn captured Sessions into automated Policy checks and proactive Slack Findings.

Everything in Developers, plus:

Automated rule generation from Sessions

5,000 QA Session reviews / mo ($0.10 / extra)

Plain-English Policy & Rule management

Proactive Slack Findings with Step evidence

Shared team workspace

Enterprise

Custom

Billed annually

Deploy inside your own cloud or EU VPC with full data sovereignty and custom controls.

Self-hosted in your AWS, GCP, or Azure VPC

Custom Session & backend capture setup

Custom data retention & access controls

Unlimited users & workspaces

Dedicated solution engineer & SLAs

Developers

$0

Free forever

Capture, inspect, and debug your production agent sessions at no cost.

Unlimited session ingestion

OpenCode, LangChain & Vercel AI SDK

1-prompt coding-agent setup

Session Explorer (Trajectory, Evidence & Graph)

Manual Findings & 1-click Fix Brief export

Softprobe MCP server (explorer.softprobe.ai/mcp)

365-day Session history

Teams

EARLY ACCESS

$499

per month, billed monthly

Turn captured Sessions into automated Policy checks and proactive Slack Findings.

Everything in Developers, plus:

Automated rule generation from Sessions

5,000 QA Session reviews / mo ($0.10 / extra)

Plain-English Policy & Rule management

Proactive Slack Findings with Step evidence

Shared team workspace

Enterprise

Custom

Billed annually

Deploy inside your own cloud or EU VPC with full data sovereignty and custom controls.

Self-hosted in your AWS, GCP, or Azure VPC

Custom Session & backend capture setup

Custom data retention & access controls

Unlimited users & workspaces

Dedicated solution engineer & SLAs

Developers

$0

Free forever

Capture, inspect, and debug your production agent sessions at no cost.

Unlimited session ingestion

OpenCode, LangChain & Vercel AI SDK

1-prompt coding-agent setup

Session Explorer (Trajectory, Evidence & Graph)

Manual Findings & 1-click Fix Brief export

Softprobe MCP server (explorer.softprobe.ai/mcp)

365-day Session history

Teams

EARLY ACCESS

$499

per month, billed monthly

Turn captured Sessions into automated Policy checks and proactive Slack Findings.

Everything in Developers, plus:

Automated rule generation from Sessions

5,000 QA Session reviews / mo ($0.10 / extra)

Plain-English Policy & Rule management

Proactive Slack Findings with Step evidence

Shared team workspace

Enterprise

Custom

Billed annually

Deploy inside your own cloud or EU VPC with full data sovereignty and custom controls.

Self-hosted in your AWS, GCP, or Azure VPC

Custom Session & backend capture setup

Custom data retention & access controls

Unlimited users & workspaces

Dedicated solution engineer & SLAs

FAQ

Questions
answered.

01. What can I use in Softprobe today?

Today, you can sign up for free at explorer.softprobe.ai, connect your OpenCode, LangChain (Python and TypeScript), or Vercel AI SDK agent with a single copy-paste prompt, stream unlimited Sessions with 365-day history, inspect step-by-step trajectories (User, Reasoning, Model call, Tool call) and end-to-end backend/API call graphs (Graph), record Findings, and export structured Fix Briefs or query live Sessions via the Softprobe MCP server in Claude Code, Cursor, or Codex.

02. How does the 1-prompt setup work?
03. How is Softprobe different from standard LLM trace viewers?
04. How do Automated Policies and Findings work in Teams Early Access?
05. Where does our agent data live, and can we self-host for GDPR compliance?

See what your agents actually do in production.

Capture every step for free. Find silent mistakes. Fix them with your coding agent.

Softprobe AI agent performance review interface preview

See what your agents actually do in production.

Capture every step for free. Find silent mistakes. Fix them with your coding agent.

Softprobe AI agent performance review interface preview

See what your agents actually do in production.

Capture every step for free. Find silent mistakes. Fix them with your coding agent.

Softprobe AI agent performance review interface preview