How AI agents track Pentagon and defense announcements
An agent tracks the Pentagon by calling a defense wire that reads the Department of Defense news release feed at runtime. It returns operations, exercises, personnel moves and official statements with timestamps and links. Contract awards are a separate question and belong on a separate wire.
/mcp. Nothing here needs an API key or a signup.
That last sentence is the part most integrations get wrong, so it is worth settling before anything else.
"What did the Pentagon say" and "who won" are different queries
The DoD news release feed is a newsroom product. A representative fetch returns items like a joint maritime exercise concluding, a task force being stood up, a National Guard wildfire deployment, a submarine finishing a forward deployment, and remarks from the Secretary. That is the shape of the source: narrative announcements about what the department is doing and saying.
It is not an award database. There is no obligation number in it, no NAICS code, no recipient DUNS or UEI, and no way to sort by dollar value. An agent that queries a news feed for "who won the tanker contract" is asking a structured question of an unstructured source, and it will either miss the award entirely or return a story that mentions it in passing.
| The question | The right route | Why |
|---|---|---|
| What did DoD announce today? | /defense/latest | News releases, statements, operations |
| Who won a federal award, for how much? | /contracts/latest | Structured award records with dollar values |
| What did the President direct? | /whitehouse/actions | Executive orders and proclamations |
If the question has a dollar sign in the answer, it is a contracts question. The routing and filters for that side are set out in a federal contract awards API for AI agents, and the practical version of the same query is in how to find the newest federal contract awards.
One feed means q is the only filter that matters
The defense wire reads a single publisher, so the src parameter accepts one value and does no useful work. Everything depends on the keyword query, and the search is ranked keyword matching rather than semantic retrieval.
That has a specific consequence worth coding around. A multi-word query has to match most of its tokens before an item is considered a candidate: a four-token query needs three of them present. So a natural-language string like "what is happening with the Navy in the Pacific" filters out almost everything, while pacific or submarine returns the items you wanted. Prefer distinctive nouns, proper names and platform names over questions.
# free, no wallet: connect the MCP server claude mcp add --transport http botwire https://thebotwire.com/mcp # paid, $0.005: DoD announcements mentioning a program, 14 day window curl -i "https://thebotwire.com/defense/latest?q=submarine&since=14d&limit=25" # a named exercise or operation, tight window curl -i "https://thebotwire.com/defense/latest?q=guard&since=48h"
Parameters are q, since (30m, 2h, 24h, 7d, 14d) and limit (1 to 50). The wire keeps a 14 day retention window and defaults to a 7 day lookback, which matters because the default is narrower than the retention. Widen since before concluding that a quiet result means a quiet week. Every route, filter and price is machine-readable at /llms-full.txt.
What an empty result actually means
Defense reporting has an unusual property: a great deal of activity is deliberately not announced, and some of what is announced arrives days after the fact. An agent that treats silence as evidence will be confidently wrong.
Three failure modes are worth guarding against explicitly:
- Operational security. Deployments and operations are frequently disclosed
after they conclude, or not at all. Absence from the feed is not absence of activity.
- Service newsrooms. Army, Navy, Air Force and combatant commands run their
own publication channels. This wire reads the department-level feed, so a service-specific release may never appear in it.
- Sourcing drift. Press coverage of defense often precedes official
confirmation. If the answer needs to be attributable, cite the announcement, not a story about it. The distinction is worked through in what primary source means for machine-readable data.
The safe output shape is the same one the other government wires use: name the announcement, give the date, link the release, and say which window was searched.
FAQ
Does this wire include the daily DoD contract announcements?
Treat it as a news feed, not a contract list. The department publishes contract announcements through its own newsroom channel, and award records land in USASpending on their own schedule. For anything where the dollar value, the recipient or the set-aside status is part of the answer, use the contracts wire instead, which returns structured award records.
Can an agent filter by military service or command?
Not as a first-class parameter, because the wire reads one department-level feed. The practical substitute is a q term for the service or command name, accepting that a service-only release published on a service newsroom will not be in scope at all. State the scope you searched rather than implying full coverage of the department.
How fresh is the defense wire?
It sits in the sparse ingest tier, polled at most every 15 minutes, which is comfortably inside the publishing rhythm of a feed that updates a handful of times per working day. The observed interval between ingest runs across the service and the last poll timestamp are both published live at /health.
Is $0.005 per call the whole cost?
Yes, settled in USDC on Base, with no API key, no signup and no minimum. A daily sweep across five watch terms is five calls a day, which is under $1 a month. The free MCP tier route returns the top three results at no cost, which is enough to test whether a query term is worth paying for.
Sources
The Department of Defense publishes its news releases as a public feed at defense.gov, which is the source this wire reads; the fetch returns up to 20 recent items. Structured federal award records, including defense awards with obligation amounts and recipient detail, are published by the government at USASpending.gov. Poll cadence and per-call pricing for this service are published live at thebotwire.com/health.
Related: A federal contract awards API for AI agents and MCP clients · How do AI agents track executive orders and presidential actions?