ContextLane

Open-source context ingestion pipeline for persistent AI agents.

MemoryLane gives agents memory. ContextLane gives agents something worth remembering.

Turn files, folders, URLs, repos, and notes into structured context with citations — then sync to MemoryLane for persistent agent recall.

Install

npm install -g @talocode/contextlane
contextlane demo

Or try without installing:

npx @talocode/contextlane demo

Quick Usage

# Ingest a file
contextlane ingest ./README.md

# Ingest a URL
contextlane ingest-url https://example.com

# Ingest a GitHub repo
contextlane ingest-github https://github.com/talocode/gatelane

# Recall what you ingested
contextlane recall "What is this project about?"

# Search across all context
contextlane search "authentication"

# Sync to MemoryLane (optional)
contextlane sync memorylane <runId>

How It Works

source → load → normalize → chunk → extract → cite → save → optional MemoryLane sync

Extracted from every run:

• Summary
• Key facts (with confidence scores)
• Decisions (active/historical)
• Action items (with priority)
• Entities (people, projects, tools, URLs)
• Tags
• Line-level citations
• MemoryLane-compatible records

Supported Sources

| Type | Examples | |------|----------| | File | .md, .txt, .json, .csv, .pdf, .ts, .js, .py, .go, .rs, .css, .html, .yaml, .sh, +40 more | | Folder | Recursive, auto-ignores node_modules/.git/dist/build/.next/coverage | | URL | Public URLs, extracts readable text from HTML | | GitHub | Clones depth 1, ingests README/docs/src | | Text | Inline text content |

CLI

contextlane --version
contextlane --help
contextlane init
contextlane doctor
contextlane ingest <path>
contextlane ingest-url <url>
contextlane ingest-github <repoUrl>
contextlane runs list
contextlane runs show <runId>
contextlane search <query>
contextlane recall <query>
contextlane sync memorylane <runId>
contextlane export <runId> --out export.json
contextlane import export.json
contextlane serve --port 3060
contextlane mcp

API

Default port: 3060

# Ingest a file
curl -X POST http://localhost:3060/v1/contextlane/ingest \
  -H 'Content-Type: application/json' \
  -d '{"input":"./README.md","type":"file"}'

# Recall
curl -X POST http://localhost:3060/v1/contextlane/recall \
  -H 'Content-Type: application/json' \
  -d '{"query":"What is this project about?"}'

SDK

import { ContextLaneClient } from '@talocode/contextlane'

const client = new ContextLaneClient({ baseUrl: 'http://localhost:3060' })
const run = await client.ingest({ input: './README.md', type: 'file' })
const recall = await client.recall({ query: 'What is this?' })

MCP

contextlane mcp

10 MCP tools for AI agents: ingest, search, recall, sync to MemoryLane, and more.

MemoryLane Sync

ContextLane generates MemoryLane-compatible records on every ingestion run.

Sync methods (in priority order):

1. HTTP → CONTEXTLANE_MEMORYLANE_URL
2. CLI → memorylane command
3. Fallback → ~/.contextlane/runs/<runId>/memory-records.json

Cloud Auth

Local usage is open-source and keyless. Talocode Cloud mode is gated by TALOCODE_API_KEY.

export CONTEXTLANE_CLOUD_MODE=true
export TALOCODE_API_KEY=your_key_here

GateLane Integration

gatelane servers add contextlane --type mcp-stdio --command "contextlane mcp"
gatelane tools discover
gatelane policy allow contextlane.contextlane_ingest
gatelane policy allow contextlane.contextlane_recall

Links

• npm: @talocode/contextlane
• GitHub: github.com/talocode/contextlane
• Docs: docs/

License

MIT