Finding signal on Twitter is more difficult than it used to be. We curate the best tweets on topics like AI, startups, and product development every weekday so you can focus on what matters.

Tech Twitter
Agent interfacepublic evidence · v1

Public evidence interface · v1

Tech Twitter for Agents

Give agents compact, citation-ready evidence about what changed, what launched, what people are arguing about, what is worth reading, and which narratives are emerging across curated tech discourse.

Endpoints

  • GET/api/agent/context

    Ranked evidence

  • POST/api/a2a

    JSON-RPC agent

Auth
None
Mode
Read only
Format
JSON
Max rows
20

Context API

GET /api/agent/context

One request, ranked evidence

Set kind=auto and pass a natural-language question, or choose a recipe directly. The response carries evidence rows, provenance, metrics, facets, instructions, and canonical citation URLs.

curl --get 'https://www.techtwitter.com/api/agent/context' \
  --data-urlencode 'kind=launches' \
  --data-urlencode 'question=What launched in developer tools?' \
  --data-urlencode 'window=7d' \
  --data-urlencode 'limit=8'

Query parameters

6 parameters
kind
enum · default auto
auto, what-changed, launches, arguments, read-list, or narrative-alert. Auto classifies the question for you.
question
string · max 500
Natural-language question. Drives auto classification and the returned recipe.
q
string · max 200
Keyword query used to rank evidence inside the window.
topic
string · max 120
Restricts evidence to a single topic name or slug.
window
enum · default 24h
24h, 48h, or 7d. Any other value is rejected as an invalid parameter.
limit
integer · default 8
Number of evidence rows to return, from 1 to 20.

Request lab

live · read only

Live GET sandbox

This form calls this deployment's read-only context endpoint with your own question. It cannot mutate data.

Live read-only sandboxcurrent deployment
Ready24h · limit 5 · ranked citations

Evidence readout

Ask the evidence index

Auto classifies the question, filters the recent evidence window, and returns ranked source records with canonical citation URLs.

window
24h
limit
5
mode
read

Authentication

no keys

No API key is required

The documented API is public and read-only. API keys are not issued. Never send Supabase service-role credentials, cron secrets, or internal job secrets to these endpoints.

Rate limits

per client

Durable public budgets

Context requests allow 120 requests per minute per client, and A2A requests allow 60. A 429 response carries Retry-After; wait that many seconds before retrying.

Boundaries

read only

Evidence, not account access

This is not a complete X firehose, and there are no posting, liking, following, webhook, user-account, or private-data operations. Cite the canonical URL on each evidence row.

Resource index

public · read-only · JSON
01OpenAPI 3.1 specificationTyped operations and schemas for function calling and generated clients./openapi.json →02llms.txt instructionsWhen to use Tech Twitter, evidence recipes, and public source links./llms.txt →03A2A agent cardMachine discovery metadata for the stateless JSON-RPC evidence agent./.well-known/agent-card.json →04Markdown agent guideA compact, agent-readable version of this page./agents.md →05Printing Press skill + CLIExternal read-only integration for offline search, local evidence, and change-over-time workflows.github · pp-techtwitter ↗

Error contract

application/json

Recover without guessing

REST failures return the same JSON envelope on every route, including unknown API paths. A2A keeps standard JSON-RPC error objects.

code
Stable machine identifier you can branch on.
message
Readable description of what the endpoint rejected.
hint
The concrete next action that resolves it.
{
  "error": {
    "code": "invalid_parameter",
    "message": "Unsupported window",
    "hint": "Use 24h, 48h, or 7d."
  }
}