Connect Tasktrail
Add the remote MCP URL, sign in with Google, and approve access. There are no loose API keys to carry around.
❯ https://tasktrail.ai/mcpone endpoint → every sessionTasktrail gives Claude Code, Codex, and other coding agents a private place to record progress, decisions, blockers, and next steps while they work.
Scaffolded the OAuth token endpoint with PKCE verification
Rotate refresh tokens on use; revoke the prior access token
Pooled connection dropped the migration write
Add a direct-connection fallback for drizzle-kit migrate
4 useful moments preserved before lunch.
OAuth gives the tools access. One standing instruction tells your agents what is worth keeping. After that, the trail grows while you stay in the work.
Add the remote MCP URL, sign in with Google, and approve access. There are no loose API keys to carry around.
❯ https://tasktrail.ai/mcpone endpoint → every sessionPut one standing instruction in AGENTS.md or CLAUDE.md, then allow the Tasktrail tools so logging stays promptless.
Use Tasktrail automatically throughout the session.
approve tools onceAgents log meaningful progress as you code. Browse the trace or ask for a deterministic day summary whenever context goes missing.
not more productivity theater—
just a faithful memory.
Claude Code and Codex log progress, decisions, blockers, and next steps while the work is still warm.
Timestamped entries make the path legible—from first attempt to final call—with the source attached.
summarize_day groups the evidence by project and task. It reports the record instead of rewriting it.
Projects and idempotent tasks keep continuing work together, so agents resume instead of duplicating.
Everything is private to your Google account. Tokens are hashed at rest and revocable at any time.
Tasktrail is a remote MCP server with OAuth, ready for Claude, Codex, and other MCP clients.
Tasktrail is a standard remote MCP server with its own OAuth. Add it in Claude’s connector settings or wire it into Claude Code and Codex from the terminal.
Open connector settings, paste the Tasktrail endpoint, then sign in with Google once.
❯ claude mcp add --transport http \ tasktrail https://tasktrail.ai/mcp ❯ claude mcp login tasktrail
❯ codex mcp add tasktrail \
--url https://tasktrail.ai/mcpThe short version: connect once, set one field rule, keep working, and come back to a private record grounded in actual traces.
Still curious? The details are in the docs ↗Sign in with Google, connect an agent, add one standing instruction, and let today’s work become tomorrow’s reliable context.