What are the Āgamas and Tantras?

There is a palm-leaf manuscript in Nepal, copied in the ninth century of the Common Era, that carries the oldest complete tantric scripture we have. It is called the Niśvāsatattvasaṃhitā, and its earliest layers were composed three or four centuries before the surviving copy was made, somewhere around the fifth or sixth century. When the editors Dominic Goodall, Alexis Sanderson and Harunaga Isaacson published it in 2015, they gave it a plain and startling subtitle: The Earliest Surviving Śaiva Tantra.

Behind that one manuscript stands a library. Not the Veda, which is older by a thousand years and more. A second library. Hindus call it the Āgamas, the Tantras, the Saṃhitās, and the tradition holds that it came down, whole and eternal, from the mouth of a god.

This is the scripture that most Hindu worship actually runs on. The mantras a temple priest recites at the consecration of an image, the geometry of a yantra, the initiation that makes a person a practitioner, the daily order of a temple’s rites: these are governed, far more often than by the Veda, by the Āgamas. And yet most people who grow up inside the tradition could not name a single one. This article is a map of that second library. Where the record is clear it says so. Where it is silent, and it is often silent, it says that too.

The word that means “what has come down”

Begin with the name, because the name is an argument.

Āgama is built from the prefix ā-, “towards, near,” and the root gam, “to go, to come.” Literally it means “a coming near, an arrival.” The Sanskrit lexicographers extend it to “a traditional doctrine, a body of such doctrine, a sacred work.” Monier-Williams glosses it exactly so. The point is that the word does not mean “book” and does not mean “revelation” in the abstract. It means that which has come down. A scripture named after the act of its own transmission is telling you, before you open it, that it does not regard itself as composed. It regards itself as received.

Set against āgama is its mirror, nigama. Where an Āgama, by the standard framing, is spoken by Shiva and heard by the Goddess, a nigama runs the other way, spoken by the Goddess and heard by Shiva. Nigama is also, confusingly and tellingly, an old synonym for the Veda itself. The two words mark the two revelations.

A verse floats through the tradition offering a folk etymology, splitting the syllables ā-ga-ma into three acts:

āgataṃ śiva-vaktrebhyo gataṃ ca girijā-mukhe / mataṃ ca vāsudevasya tasmād āgama ucyate //

“Come down from Shiva’s mouths, gone into the mouth of Girijā the mountain’s daughter, and approved by Vāsudeva: therefore it is called Āgama.” Come, gone, approved. Āgatam, gatam, matam. It is a pun more than a derivation, and its home is unstable. The Goudriaan-Gupta catalogue of tantric literature reports it from the Piṅgalāmata; other recensions give it to the Paramasaṃhitā, others to the Svacchanda Tantra. It should not be assigned to any single scripture with confidence, and it should not, despite a common claim online, be assigned to the Kulārṇava Tantra. What matters is not its address but its content. The tradition’s own account of its scripture is an account of a handing-down.

One revelation, three streams

The second library is not one river. It is three, and keeping them apart is the first discipline of reading it.

The division is by the deity who does the revealing. Where Shiva reveals, the scripture is usually called an Āgama, and the stream is Śaiva. Where the Goddess is the centre, it is usually called a Tantra, and the stream is Śākta. Where Vishnu or Narayana reveals, it is called a Saṃhitā, “a collection,” and the stream is Vaiṣṇava, its corpus known as the Pāñcarātra. Three deities, three names, three canons.

The naming convention is real but soft. It is a tendency, not a law. Some of the most important Śaiva scriptures wear the word Tantra in their titles, the Svacchanda, the Netra, the Mālinīvijayottara. Pāñcarātra texts are also called Āgamas. André Padoux, whose The Hindu Tantric World is the best short survey in English, simply lists tantra, āgama, saṃhitā, sūtra together as the near-interchangeable names of a single kind of literature. So when the Śaiva is called Āgama and the Śākta is called Tantra, read it as habit, not rule.

Each stream regards its own scripture as complete and, as a tendency, does not treat the others’ canons as binding. The Śaiva does not derive authority from the Vaiṣṇava, nor the Vaiṣṇava from the Śākta. This is a tendency and not an absolute: shared ritual grammar and later digest-literature cross the sectarian lines freely. But the working assumption of each tradition is that its own god revealed the sufficient scripture.

Within these streams sit the traditions this site treats elsewhere. The Śaiva splits, in Sanderson’s map, into the dualist ritual mainstream, the Śaiva Siddhānta, and the more esoteric, Goddess-inflected Bhairava tantras above it. The Śākta divides into the fierce Kālīkula, the family of Kali, and the benign Śrīkula, the family of Tripurasundari and the Śrīvidyā. The three streams are treated one by one in the companion articles: the 28 Śaiva Āgamas, the 64 Śākta Tantras, and the Pāñcarātra Saṃhitās.

The dialogue of Shiva and the Goddess

Open almost any of these scriptures and you find the same scene. The Goddess asks. Shiva answers. The doctrine unfolds as a conversation between them, question and reply, on a mountain outside of time.

This is the revelation frame, and it is doing theological work. The teaching is not a monologue delivered downward. It is drawn out by a questioner, and the questioner is the God’s own consort, his śakti, his power. When the direction reverses, when Shiva is made to ask and the Goddess to reveal, the text is often classed as a nigama rather than an āgama, and the Rudrayāmala is the stock example. A verse in the Svacchanda Tantra, preserved by Sir John Woodroffe, resolves the apparent puzzle of who really speaks by collapsing the two roles into one: “Standing himself in the place of both guru and disciple, the Great Lord brought down the Tantra by way of question and answer.” The dialogue is a device. The one consciousness plays both parts.

The four quarters of a scripture

An Āgama, in its ideal form, is built in four quarters, four pādas, and each governs a department of the religious life.

Jñāna-pāda, the quarter of knowledge: metaphysics, the levels of reality, the analysis of the soul and its bondage. Yoga-pāda, the quarter of discipline: meditation and the inner practices. Kriyā-pāda, the quarter of action: the vast technical apparatus of ritual, the building of temples, the making of images, the rites of consecration. Caryā-pāda, the quarter of conduct: daily observance, festivals, the rules that govern the various grades of initiate. Doctrine, meditation, ritual, conduct. It is an elegant scheme, and Sanderson reports it as the conventional Śaiva division of the means of liberation.

Now the honest correction, which the specialist Hélène Brunner-Lachaux insisted on. The four-pāda scheme is an ideal, not a description. Most of the surviving Āgamas are not actually divided into four such sections at all. They are overwhelmingly weighted toward kriyā and caryā, the ritual and the conduct, while the jñāna and yoga quarters are frequently thin, late, or simply missing. The tradition drew a fourfold map. The territory it mapped is lopsided toward ritual. Both facts are true, and a source-faithful account holds them together instead of choosing the tidy one.

How old, and how we date the undatable

The tradition’s answer to “how old” is that the question does not apply. The Āgamas are held to be beginningless, revealed and re-revealed in age after age. The historian’s answer is different, and it is hard-won, because the corpus is designed to resist dating.

These texts are anonymous. They are composite, grown in layers over centuries, with old titles retained over new material. They carry no colophon that says who wrote them or when. So the relative chronology has to be built sideways, from citation networks, from dated manuscripts, from inscriptions that mention a scripture by name. Sanderson’s long essay “The Śaiva Literature” is, in effect, one such map, ordering the corpus by who quotes whom.

What emerges is a literature coming into the historical record from roughly the fifth or sixth century of the Common Era, with the Niśvāsa at the early edge, and reaching its great productive height between about the fifth and the thirteenth centuries. Sanderson gave that long period a name that has stuck: the Śaiva Age. His argument, and it is an argument, a scholar’s thesis rather than a settled fact, is that during those eight hundred years Śaivism and its tantric literature were the dominant religious idiom of the subcontinent, shaping royal temples, land grants, kingship itself, and lending ritual forms even to Buddhism and to the Vaiṣṇava Pāñcarātra. The Veda never went away. But the working scripture of the medieval Indian temple was increasingly the Āgama.

A revelation for the present age

Why, if the Veda already stood, did a second revelation come at all? The tradition’s answer is temporal. This is the Kali Yuga, the last and most degenerate of the four ages, and the older Vedic means no longer bite. The Āgama presents itself as the revelation fitted to the present age, direct and efficacious where the Veda has grown remote. The Kulārṇava Tantra states the doctrine as a rule of ages: the practice of the first age was drawn from śruti, of the second from smṛti, of the third from the Purāṇas, and of this last age from the Āgama alone.

The Śaiva went further than fitness for the age. It positioned its scripture as a viśeṣaśāstra, a special or higher teaching, above the general Vedic dispensation, valid but surpassed. That claim, and the fierce orthodox reply to it, is a whole story of its own, the story of whether a second revelation can be revelation at all. It is told in the companion article, Āgama and Veda: two revelations.

For now it is enough to have the shape of the thing. A second library, held to be eternal, datable to the medieval centuries. Three streams, sorted by their gods. A scripture that arrives as a dialogue and is built, in theory, in four quarters. The Veda is heard. The Āgama comes down.

The tradition does not need this map. It has kept its scriptures alive for fifteen centuries without one. But for a reader standing outside the temple wall, wondering what the priest is reciting and where the words come from, the map is a place to begin.