v1.18.6May 20, 2026
Toolbelt for AI Agents 🧰
Toolbelt for AI Agents 🧰. This release brings HTTP Client for AI Agents, Database for AI Agents, Auto-Detect Installed CLIs, Sub-Agent Badge, History Viewer and OpenCode Joins MCP Settings.
Download v1.18.6
The download page always has the newest build; installers for this version are on GitHub.
What changed
New features
14- AI agents can fire HTTP requests through 1DevTool's HTTP client — Claude, Codex, Gemini, and OpenCode now get a new request MCP tool that runs any HTTP request through your active project's environment (bearer tokens, basic auth, headers, body) and returns status, headers, body, and latency — no need to paste curl commands into chat
- Replay saved requests by name — A run_saved_request tool lets agents fire any request from your HTTP collection using its saved environment and headers, so you can ask "re-run the staging health check" and the agent picks the right request, base URL, and auth on its own
- Discover what's available from the AI side — list_environments and list_saved_requests let agents browse your HTTP collection and active environment without you having to paste URLs every time
- AI agents can query your configured database connections — A new 1devtool-database MCP server exposes Postgres, MySQL, MongoDB, Redis, and every other engine 1DevTool supports to AI agents. They call list_connections first (no credentials ever leave 1DevTool), then describe_schema, preview_table, or query
- AI query results open as a tab in the Database panel — When an agent runs a query or previews a table, the result streams into a new "AI query" tab next to your own queries, with the SQL captured into history so you can audit exactly what the agent asked and what it saw
- Existing safety rules still apply — Agent-issued queries respect each connection's safeMode flag and use the same read-only / write detection as the manual SQL editor, so an agent can't bypass a read-only connection
- New "Detected CLIs" section under Settings → General → System — Lists every supported AI agent (Claude, Codex, Gemini, OpenCode, Amp, Qwen, Aider) plus runtimes, package managers, dev tools, and database CLIs that 1DevTool can find on your PATH, so "is this on my system?" stops being a guessing game
- Per-row state with real diagnostics — Each CLI shows detected, override, ambiguous (multiple paths found — pick one), not-found, unverified, timeout, or error, with the actual paths, detected version, and scan duration. Slow paths get quarantined automatically with an "Include anyway" escape hatch
- Cancellable scan with per-row Retry — Scans expose a Cancel button after 4 seconds and per-row Retry if the global timeout trips, so a flaky NFS share or a broken symlink in your PATH can't lock the dialog up
- Custom CLIs — Add your own CLI to the detected list with a name and binary, and it shows up alongside the built-ins everywhere 1DevTool reaches for a CLI
- Sub-agent badge on running terminals — When an AI CLI spawns a nested sub-agent (e.g. a Claude Bash sub-agent or a tool-call agent), a small badge appears on the terminal showing which CLI is running underneath and how long it's been alive
- Click the badge to open the Sub-Agent History viewer — A separate window live-tails the sub-agent's session log, color-coded per CLI (Claude amber, Codex green, Gemini violet, Amp pink, OpenCode blue, Qwen indigo, Aider red), with auto-scroll, file-size tracking, and incremental tail polling
- Works across every supported CLI's session format — JSONL, JSON, and markdown logs are parsed the same way, surfacing assistant text, thinking, tool calls, tool results, system messages, and turn durations with the labels each CLI actually uses
- OpenCode added alongside Claude, Codex, and Gemini in Settings → MCP — One-click install / uninstall of 1DevTool's MCP servers (channels, database, design, http) into ~/.config/opencode/opencode.json, matching the existing flow for the other clients
Improvements
7- Drag-and-drop OS files directly onto a sub-folder at any depth — Previously dragging files from Finder / Explorer always landed them at the project root; now hovering any folder highlights it as the drop target and dropping there copies the files into that folder
- Auto-expand on drop — Dropping into a collapsed sub-folder also expands it so you immediately see the files that just landed instead of having to hunt for them
- Opening a file from the file tree now activates the editor panel in whichever slot it lives — If the editor tab wasn't selected (e.g. you had the Browser or Database panel showing in the center), the editor now jumps to the front so the file you just opened isn't hidden behind another tool
- Two sub-tabs: "Tools" and "External clients" — The Tools sub-tab lists every MCP tool registered with 1DevTool — name, profile, description, output kind, mutates / long-running flags, enable state — so you can see exactly what's exposed to AI agents at a glance
- Bridge health shows registry version and tool count — Next to the bridge port, the health badge now reports how many tools are registered and the registry's revision, so a stale process or a bridge restart is immediately obvious
- Browser viewport dimensions are tracked live, not snapshotted — The Send-to-Terminal dialog now reflects your current panel size, updating whenever you resize the panel, toggle the debug console, or switch device-emulation modes, instead of capturing the size at the moment the dialog first opened
- Terminal session recording removed — The "record terminal session" / MP4 export / share-menu flow has been retired in this release; transcript copy/paste, the new Sub-Agent History viewer, and channel history cover the same ground with less footprint. Recording will return in a different form if there's demand for it