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.
npm install -g @talocode/contextlane
contextlane demopip install contextlane
contextlane demo# 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>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
| 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 |
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 mcpDefault 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?"}'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?' })from contextlane import ContextLaneClient
client = ContextLaneClient('http://localhost:3060')
run = client.ingest('./README.md')
results = client.search('What is this?')contextlane mcp10 MCP tools for AI agents: ingest, search, recall, sync to MemoryLane, and more.
ContextLane generates MemoryLane-compatible records on every ingestion run.
Sync methods (in priority order):
- HTTP →
CONTEXTLANE_MEMORYLANE_URL - CLI →
memorylanecommand - Fallback →
~/.contextlane/runs/<runId>/memory-records.json
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_heregatelane servers add contextlane --type mcp-stdio --command "contextlane mcp"
gatelane tools discover
gatelane policy allow contextlane.contextlane_ingest
gatelane policy allow contextlane.contextlane_recall- npm: @talocode/contextlane
- PyPI: contextlane
- GitHub: github.com/talocode/contextlane
- Docs: docs/
Part of Talocode — open-source workflow layers for builders. Explore sibling projects:
| Project | What it is |
|---|---|
| ScreenLane | Screen-aware voice command layer |
| Tera | AI chat & assistant |
| Codra | Local coding agent |
| GateLane | MCP gateway & agent tool control plane |
| ContextLane | Context ingestion for persistent agents (this repo) |
| MemoryLane | Persistent agent memory |
| SignalLane | X growth intelligence |
| ReplyLane | X reply opportunity intelligence |
| CrawlerLane | Crawler / SEO intelligence |
| WebDataLane | Web extraction to structured data |
| SearchLane | Search layer for agents |
| InvoiceLane | Invoicing tools |
| GeoLane | Geo intelligence |
| UgcLane | UGC workflows |
| OpenSourceLane | Open-source distribution tools |
| StackLane | Builder stack platform |
| Tradia | Trading intelligence |
| Agent Browser | Browser automation for agents |
| Talocode | Org home & control plane |
| Skills | Shared agent skills |
| X Agent | X automation agent |
| LaunchPix | Launch tooling |
| ForgeCAD | CAD workflows |
| WorkLane | Work automation |
| ClipLoop | Clip / video loops |
MCP-compatible agents integrate via each product's MCP server where available (Model Context Protocol).
More: github.com/talocode · talocode.site · docs.talocode.site
MIT